null  User manual
TRAINSCOPE
MOBILE VIDEO FOR TRANSIT SECURITY
SYSTEM DESIGN DOCUMENT
Revision 2.1
December 5, 2005
CMU-CyLab-05-002
Carnegie Mellon University
CyLab - Visual Intelligence Studio, CIC-2218
4720 Forbes Ave.
Pittsburgh, PA 15213
Contact: Dr. Yang Cai <[email protected]>
This document is delivered as-is. The authors and Carnegie Mellon University are not
responsible for any damages or losses. Please do not redistribute this document.
1
TABLE OF CONTENT
1
SYSTEM OVERVIEW
5
2
BACKGROUND
6
3
DIGITAL NETWORK CAMERAS
7
3.1
Camera architecture
7
3.2
Digital-Analog comparison
8
4
WIRELESS NETWORK 802.11G
4.1
Advantages of 802.11g
4.2
Wi-Fi security
5
DYNAMIC BANDWIDTH MANAGEMENT
9
9
10
11
5.1
Bandwidth computation
11
5.2
Network Bandwidth
12
5.3
TrainscopeViewer
14
6
MULTI-POINT ACCESSING
17
6.1
Network design
17
6.2
Signal Handover
18
6.3
Multiple bridge access points for multiple trains
21
7
DISTANCE ISSUES
23
8
VIDEO QUALITY
26
9
SCALABILITY
28
9.1
Storage
28
9.2
Network Hardware
28
9.3
Pittsburgh Technology Center test for throughput
34
9.4
Pittsburgh Technology Center test for handover
38
9.5
Pittsburgh Technology Center test for multiple bridges
41
9.6
Bombardier Transportation Test
45
ANNEXE 0 - GLOSSARY OF ACRONYMS
50
ANNEXE 1 – HARDWARE LIST SUMMARY
51
ANNEXE 1A - LINKSYS WET54GV2 DATASHEET
53
ANNEXE 1B – NETGEAR SWITCH 5 PORTS
55
ANNEXE 1C – AXIS 210 DIGITAL CAMERA
56
ANNEXE 1D – CISCO CATALYST 3500 SERIES XL
57
ANNEXE 1E - CISCO AIRONET 1200 SERIES
58
2
ANNEXE 2 : ANTENNAS SPECIFICATIONS
61
Tables
TABLE 1 : DIGITAL AND ANALOG COMPARISON TABLE
8
TABLE 2 : 802.11X COMPARISON TABLE
9
TABLE 3 : WIMAX VS. WI-FI
10
TABLE 4 : HTTP PARAMETERS FOR DYNAMIC MANAGEMENT
15
TABLE 5 : CONFIGURATION OF THE ACCESS POINT TO AVOID HANDOVERS
19
TABLE 6 : NETWORK RANGE FOR AIRONET 1200 ACCESS POINT (CISCO)
24
TABLE 7 : NETWORK CONFIGURATION FO SEVERAL NETWORK RANGE*
32
TABLE 8 : CONFIGURATION OF THE ACCESS POINTS (ACCESS POINT 1 & 2)
36
TABLE 9 : NETWORK MEASURES FOR THE PITTSBURGH TECHNOLOGY CENTER
37
TABLE 10 : RESULTS FOR THE PITTSBURGH TECHNOLOGY CENTER
38
TABLE 11 : CONFIGURATION OF THE ACCESS POINTS FOR A THRESHOLD EQUAL TO 11MBPS
39
TABLE 12 : RESULTS FOR THE PITTSBURGH TECHNOLOGY CENTER – HANDOVER TEST
40
TABLE 13 : RESULTS FOR THE PITTSBURGH TECHNOLOGY CENTER – MULTIPLE BRIDGES TEST1
43
TABLE 14 : RESULTS FOR THE PITTSBURGH TECHNOLOGY CENTER – MULTIPLE BRIDGES TEST2
44
TABLE 15 : RESULTS FOR BOMBARDIER TRANSPORTATION TEST
48
3
Figures
FIGURE 1 : TRANSIT AIRPORT SHUTTLE FROM BOMBARDIER TRANSPORTATION ............................................. 6
FIGURE 2 : NETWORK CAMERA ARCHITECTURE ............................................................................................... 7
FIGURE 3 : IMAGE SIZE DIAGRAM .................................................................................................................. 11
FIGURE 4 : DYNAMIC BANDWIDTH SCENARIO ................................................................................................ 12
FIGURE 5 : 38MBT/S USED FOR A STATIC MANAGEMENT (100%=100MBT/S) ................................................ 13
FIGURE 6 : 38MBT/S – 9MBT/S FOR A BASIC DYNAMIC BANDWIDTH MANAGEMENT (100%=100MBT/S) ...... 13
FIGURE 7 : EVOLVED DYNAMIC BANDWIDTH MANAGEMENT 22MBT/S – 8MBT/S .......................................... 13
FIGURE 8 : HTTP REQUEST TO CHECK THE STATUS OF THE CAMERAS OF THE NETWORK ............................... 14
FIGURE 9 : HTTP REQUEST TO MODIFY RESOLUTION AND COMPRESSION ...................................................... 14
FIGURE 10 : HTTP REQUEST TO MODIFY THE FRAME PER SECOND ................................................................ 14
FIGURE 11 : ONBOARD NETWORK DIAGRAM FOR TRAINET......................................................................... 17
FIGURE 12 : LAND NETWORK DIAGRAM FOR TRAINET................................................................................ 17
FIGURE 13 : NETWORK DESIGN FOR ACCESS POINT COLLABORATION ............................................................ 18
FIGURE 14 : MULTIPLE BRIDGE ACCESS POINTS FOR MULTIPLE TRAINS ......................................................... 21
FIGURE 15 : NETWORK TRAFFIC VS. DISTANCE .............................................................................................. 23
FIGURE 16 : MAXIMUM BANDWIDTH FOR OUTDOOR CONDITIONS 26 MBTS/S ................................................ 23
FIGURE 17 : LINKING QUALITY VS. DISTANCE FROM ACCESS POINTS ........................................................... 24
FIGURE 18 : COVERAGE OF THE OMNI DIRECTIONAL ANTENNA ..................................................................... 24
FIGURE 19 : IMAGE QUALITY FOR 640X480 WITH BACKLIGHT COMPENSATION ............................................. 27
FIGURE 20 : BANDWIDTH VS. NUMBER OF CAMERA PER VEHICLE AT 30 FPS FOR DIFFERENT RESOLUTION .... 29
FIGURE 21 : BANDWIDTH VS. NUMBER OF CAMERA PER VEHICLE AT 15 FPS FOR DIFFERENT RESOLUTION .... 30
FIGURE 22 : BANDWIDTH VS. NUMBER OF CAMERA PER VEHICLE AT 30 FPS FOR DIFFERENT COMPRESSION.. 31
FIGURE 23 : BANDWIDTH VS. NUMBER OF CAMERA PER VEHICLE AT 15 FPS FOR DIFFERENT COMPRESSION.. 31
FIGURE 24 : BANDWIDTH VS. NUMBER OF VEHICLE AT 15 FPS 320X240 30% ............................................... 32
FIGURE 25 : NUMBER OF BRIDGES VS. NUMBER OF VEHICLES (4 CAMERAS OR 6 CAMERAS) AT 30 FPS .......... 33
FIGURE 26 : NUMBER OF BRIDGES VS. NUMBER OF VEHICLES (4 CAMERAS OR 6 CAMERAS) AT 15 FPS .......... 34
FIGURE 27 : ONBOARD NETWORK DIAGRAM FOR PTC TEST ........................................................................ 34
FIGURE 28 : LAND NETWORK DIAGRAM FOR PTC TEST ............................................................................... 35
FIGURE 29 : CAMERA SET UP ......................................................................................................................... 36
FIGURE 30 : SATELLITE VIEW PITTSBURGH TECHNOLOGY CENTER TEST ...................................................... 37
FIGURE 31: LAND NETWORK DIAGRAM FOR PTC TEST ................................................................................ 38
FIGURE 32 : SATELLITE VIEW PITTSBURGH TECHNOLOGY CENTER TEST ...................................................... 39
FIGURE 33 : BANDWIDTH VS. DISTANCE (THRESHOLDS 11 MBPS, 12 MBPS AND 18 MBPS) ........................... 40
FIGURE 34 : ONBOARD NETWORK FOR VEHICLE 1 ......................................................................................... 41
FIGURE 35 : ONBOARD NETWORK FOR VEHICLE 2 ......................................................................................... 41
FIGURE 36 : CAMERA SCREENSHOTS FOR MULTIPLE BRIDGES SCENARIO 1 .................................................... 42
FIGURE 37 : MULTIPLE BRIDGES WITH TWO VEHICLE FOLLOWING EACH OTHER ............................................ 42
FIGURE 38 : CAMERA SCREENSHOTS FOR MULTIPLE BRIDGES SCENARIO 1 .................................................... 43
FIGURE 39 : MULTIPLE BRIDGES WITH TWO VEHICLES INTERCEPTING ........................................................... 44
FIGURE 40 : BOMBARDIER TRANSPORTATION TEST TRACK SATELLITE VIEW................................................. 45
FIGURE 41 : SHUTTLE VEHICLE ON THE TEST TRACK ..................................................................................... 45
FIGURE 42 : ONBOARD NETWORK DIAGRAM FOR TRAINET......................................................................... 46
FIGURE 43 : LAND NETWORK DIAGRAM FOR TRAINET................................................................................ 46
FIGURE 44 : IMAGE RECORDED THROUGH THE WIRELESS NETWORK .............................................................. 47
FIGURE 45 : NETWORK TRAFFIC FOR BOMBARDIER TEST TRACK (AVERAGE 8 MBPS) (100% 100MBPS) ...... 47
4
1 System Overview
The purpose of this project is to develop a concept-proofing real-time video surveillance
technology for security of transit vehicles. Our objective is to investigate the Quality of
Service (QoS), affordability and scalability of the mobile video system. Our approach is
the open-system design, which is affordable and flexible. The system includes
inexpensive off-the-shelf digital cameras on board and a mobile network connected to a
land network through a wireless connection. The Dynamic Bandwidth Management
software was developed to expand the bandwidth for the wireless transmission of
streaming video. The project was carried out by Carnegie Mellon University, with
partnership with Bombardier Transportation. Our tasks for this project include:
•
•
•
•
•
Architecture of a multi-point wireless assessing network video system
Development of the dynamic bandwidth management model and software
Quality of Service (QoS) research for the relationship of video spatiotemporal
resolution, wireless bandwidth and distance, et al.
Scalability studies for multiple camera and multiple vehicle configurations
Applications of the mobile video, e.g. unattended baggage detection, etc.
From the field tests in Bombardier Transportation West Mifflin Test Track on July 27,
2005, the investigators found that the system design matches the criteria for a CCTV
system. The video flow from the four cameras was transmitted wirelessly in real time.
The dynamic bandwidth management used only 30% of the connection capacity for a
single vehicle.
The tests showed that the ‘line-of-sight’ of the wireless signal is more important than the
distance, which was the original concern. This finding is useful in designing the wireless
antenna and determining the locations of the wireless repeaters. The test also showed that
a proper handover configuration can improve the throughput of the wireless network.
The investigators have also developed additional systems based on the mobile video,
such as the unattended baggage detection, etc. (in a separate Cylab Technical Report).
These prototypes have been tested with realistic samples such as video from the inside of
the airport shuttles (see attached appendix).
5
2 Background
Terrorist attacks at transit systems have increased worldwide during the past decade.
Forty percent of terrorist targets worldwide in 1998 were transportation targets, with a
growing number against transit systems. The tragic event of September 11th, Madrid in
2004 and the bombings in the subways and buses of London in 2005 along with the
continuous threat of terrorism prompted U.S. transit agencies to take stock of their
security procedures and prepare for more life threatening situations.
Figure 1 : Transit airport shuttle from Bombardier Transportation
6
3 Digital Network Cameras
3.1 Camera architecture
A network camera is fully bi-directional and also integrates with the rest of the system to
a high degree in a distributed and scalable environment. This is a contrast to an analog
camera that is a one-directional signal carrier which terminates at the DVR and operator
level. A network camera communicates with several applications in parallel, to perform
various tasks, such as detecting motion or sending different streams of video.
A network camera can be described as a camera and computer combined into one unit.
It connects directly to the network as any other network device. A network camera has its
own IP address and built-in computing functions to handle network communication.
Everything needed for viewing images over the network is built into the unit. A network
camera has built-in software for a Web server, FTP server, FTP client and e-mail client.
Other features include alarm input and relay output functions.
Figure 2 : Network camera architecture
The network camera's camera component captures the image -- which can be described as
light of different wavelengths -- and transforms it into electrical signals. These signals are
then are converted from analog to digital format and transferred into the computer
function where the image is compressed and sent out over the network. The lens of the
camera focuses the light onto the image sensor (CCD/CMOS). Before reaching the image
sensor, the images pass through the optical filter, which removes any infrared (IR) light
so that the "correct" colors will be displayed. (In day/night cameras, this IR-cut filter is
removable to provide high quality black & white video during nighttime conditions.) The
image sensor converts the image, which is composed of light information, into electrical
signals. These electrical, digital signals are now in a format that can be compressed and
transferred over networks.
7
3.2 Digital-Analog comparison
A digital network camera-based
system
Access
As open or closed access as needed. Remote access to
live images and remote administration of a network
camera are possible from anywhere using a standard
Web browser on any PC.
- You can administer and view the images remotely
using a standard Web browser on any PC.
Closed circuit. No possibility for remote access.
- Images can be recorded on a hard disk, enabling easy
search possibilities, easy storage and no image
degradation or wear.
- Images must be stored on video tape cassettes,
which require constant changing and lots of
storage space. The quality of recorded images
deteriorate over time.
Ease of use
- The hard disk can be located at a remote location for
security purposes.
Quality
System
requirements
Installation
Cabling
Scalability
An analog camera-based
system
Digital images do not lose quality in transmission or
storage. A digital picture is created using Motion-JPEG.
Once created, the image is free from degradation. Each
frame within a video stream is sharp.
Everything needed to stream live video over networks is
included in the network camera. Simply connect the
network camera to a network. View, record and
administer from any networked PC (located anywhere).
Simply connect a network camera to the nearest
network connection and assign an IP address.
One standard UTP (unshielded twisted pair) network
cable can forward images from hundreds of network
cameras simultaneously.
Adding more network cameras to the system is easy.
A high quality network cable typically costs 30 to 40
percent less than a standard coaxial cable.
A network cable can also support hundreds of network
cameras and other devices.
Cost
An IP-based network infrastructure is often already in
place, which means the cost is reduced to only that of
the network camera(s).
- Remote administration or monitoring is not
possible.
- The video cassette recorder must be located
near the camera. This could potentially enable
unauthorized persons to have access to the video
tape.
Image quality is lost when using long cables and
the resolution of a magnetic tape is normally
quite low. In addition, the quality of the recorded
video deteriorates over time.
Connection to a coaxial cable, to a multiplexer, to
a video or time lapse recorder, and to a locally
placed CRT (cathode ray tube) monitor.
Attach a coax cable to each and every camera
and connect to the multiplexer.
One cable can transport video signals from only
one camera at a time. If you have two cameras,
you have to have two cables. This often means
large cable trunks filled with thick and sensitive
cables that are connected to a locally placed
control room.
Very difficult. Each analog camera requires its
own cable. Image quality is lost when using long
cables.
Expensive coaxial cables. A classic RG59 75
Ohms coaxial cable typically costs 30 to 40
percent more than a high quality network cable.
In addition, more cable is required. Each analog
camera requires its own cabling. High labor and
maintenance demands, plus cost of the analog
camera(s), video tape recorder and video tape
cassettes.
Table 1 : Digital and analog comparison table
8
4 Wireless Network 802.11g
4.1 Advantages of 802.11g
In 2002 and 2003, Wireless Local Area Network products supporting a new standard
called 802.11g began to appear on the scene. 802.11g attempts to combine the best of
both 802.11a and 802.11b. 802.11g supports bandwidth up to 54 Mbps, and it uses the
2.4 Ghz frequency for greater range. 802.11g is backwards compatible with 802.11b,
meaning that 802.11g access points will work with 802.11b wireless network adapters
and vice versa.
•
Pros of 802.11g - fastest maximum speed; supports more simultaneous users;
signal range is best and is not easily obstructed and not expensive.
Standard
Maximum Bit Rate
Fallback Rates
Channels Provided
Frequency Band
Radio Technique
802.11
2 Mbps
1 Mbps
3
2.4 GHz
FHSS or DSSS
802.11b
11 Mbps
3
2.4 GHz
DSSS
12
5 GHz
OFDM
3
2.4 GHz
OFDM
5.5 Mbps
2 Mbps
1 Mbps
48 Mbps
36 Mbps
24 Mbps
802.11a
54 Mbps
18 Mbps
12 Mbps
6 Mbps
2 Mbps
802.11g
54 Mbps
Same as 802.11a
Table 2 : 802.11x comparison table
One approach to increasing the physical transfer rates of wireless systems employs
multiple antenna systems for both the transmitter and the receiver. This technology is
referred to as multiple-input multiple-output (MIMO), or smart antenna systems. MIMO
exploits the use of multiple signals transmitted into the wireless medium and multiple
signals received from the wireless medium to improve wireless performance.
MIMO can provide many benefits, all derived from the ability to process spatially
different signals simultaneously. Two important benefits explored here are antenna
diversity and spatial multiplexing. Using multiple antennas, MIMO technology offers the
ability to coherently resolve information from multiple signal paths using spatially
separated receive antennas. Multipath signals are the reflected signals arriving at the
receiver some time after the original or line of sight (LOS) signal has been received.
Multipath is typically perceived as interference degrading a receiver's ability to recover
the intelligent information. MIMO enables the opportunity to spatially resolve multipath
signals, providing diversity gain that contributes to a receiver's ability to recover the
9
intelligent information. While WiMax has a wider range but it’s multipoint devices are
not as mature as 802.11g.
Primary development
Frequency Band
Channel bandwidth
Bandwidth efficiency
Half/Full duplex
Modulation
FEC
Encryption
Access Protocol
Best Effort
Data Priority
Consistent Delay
Radio Technology
WiMax
(802.16a)
Broadband wireless access
Licensed/Unlicensed
2G to 11 GHz
Adjustable
1.25 M to 20 MHz
Wi-Fi
(802.11g)
Wireless LAN
2.4 GHz ISM (g)
< 5bps/Hz
Full
BPSK, QPSK
16-,64-, 256-QAM
Convolutional Code
Reed-Solomon
Mandatory – 3DES
Optional AES
Request/Grant
Yes
Yes
Yes
OFDM (256 channels)
< 2.7 bps/Hz
Half
BPSK, QPSK
16-,64-QAM
Convolutional Code
20 MHz
Optional AES
CSMA/CA
Yes
802.11e WME
802.11e WSM
OFDM (64 channels)
Table 3 : WiMax vs. Wi-Fi
4.2 Wi-Fi security
In Wi-Fi, encryption is optional, and three different techniques have been defined:
¾ Wired Equivalent Privacy (WEP): An RC4-based 40- or 104-bit encryption
with a static key.
¾ Wi-Fi Protected Access (WPA): A new standard from the Wi-Fi Alliance that
uses the 40- or 104-bit WEP key, but changes the key on each packet to thwart
key- crackers. That changing key functionality is called the Temporal Key
Integrity Protocol (TKIP).
¾ IEEE 802.11i/WPA2: The IEEE is finalizing the 802.11i standard, which will
be based on a far more robust encryption technique called the Advanced
Encryption Standard. The Wi-Fi Alliance will designate products that comply
with the 802.11i standard as WPA2. However, implementing 802.11i will
typically require a hardware upgrade, so while the standard should be
completed in mid-2004, it might be some time before it is widely deployed.
10
5 Dynamic Bandwidth Management
5.1 Bandwidth computation
For wireless network the bandwidth is the limiting factor and displaying more than 10
video with high resolution, high frame rate and high quality is currently a big issue.
Typically, for a set up of four cameras capturing 640*480 video at 30 fps with a
compression rate of 10%, the bandwidth required is 36.4 Mbit/sec. This is the reason why
a system with more than 12 cameras overtakes the capacity of traditional Ethernet cable
(100 Mbits/sec).
In practice, a video camera displayed on the small screen with a high resolution does not
provide a significant improvement for the viewer and is a waste of bandwidth. The
correlation between image size, compression and resolution level is illustrated
schematically in Fig. 3. The size of the picture from the motion JPEG flow increases with
the resolution and the quality (less compression).
Image Size function of compression and resolution
40
Image Size
(Ko)
30
30-40
20-30
10-20
0-10
20
10
640*480
0
10
320*240
30
50
Compression Rate (% )
70
90
160*120
Image
resolution
Figure 3 : Image size diagram
A model of this function is created in order to optimize the link between human
perception and image size. We obtained three equations (1)(2)(3) to compute the
bandwidth. The exponential equation (2) is linked to the nature of JPEG compression.
JPEG is designed for compressing full-color or gray-scale images of natural, real-world
scenes. JPEG removes redundant information from images which is why equation (2) is
based on the exponential: the non-redundant information could not be totally removed
and remains as the lowest size of the picture. The image resolution I in pixels and the
compression rate percentage CR are the main parameters used to compute the image size
S in pixels and the resulting bandwidth B (Mbits per second). First, the non-redundant
information RI can be approximated by equation (1).
11
RI = 3.44*10− 4 * I + 7.41
(1)
S = RI * (CR) −0.40
(2)
B = S * CN * FPS * SC
(3)
The constant SC represents the complexity of a scene. This constant is in most instances
equal to 8. It is used along with the number of cameras CN and the frames per seconds
fps for each camera to carefully estimate the bandwidth.
5.2 Network Bandwidth
To demonstrate the performance of our system, we compare our CMU-system to existing
systems. We have established some real scenarios to analyze the bandwidth adaptation
when switching between different views. First, the user is watching a display with 4 equal
screens. The screen selected is displayed in a full screen mode for 10 seconds. Then the
user is returned to the four screen view. This scenario (Fig. 4) is tested with default static
parameters: resolution (640x480) and compression (10%). For the dynamic management,
these two parameters will change according to the human behavior. As explained
previously, the number of frames per second is maintained at 30 to guarantee a full
movement perception.
Figure 4 : Dynamic bandwidth scenario
The first comparative test is based on static bandwidth management. The default
resolution (640x480) and compression (10%) are selected permanently. These parameters
are chosen in order to obtain high definition video when the user selects a screen in a full
screen mode. The results (Fig. 5) were predictable. Because none of the parameters
evolve, the bandwidth does not change and 38Mbt/s is used permanently to display 4
cameras in high definition.
12
Figure 5 : 38Mbt/s used for a static management (100%=100Mbt/s)
The second test is based on basic dynamic management of the bandwidth. When the user
switches to the full screen mode, the bit flow from the three other video cameras is
disabled. It results a rectangle signal (Fig. 6) with the high level corresponding to the four
screen mode (38Mbt/s) and the low level to the full screen mode (9Mbts). However, the
bandwidth used globally is still too large to support over 12 video cameras.
Figure 6 : 38Mbt/s – 9Mbt/s for a basic dynamic bandwidth management (100%=100Mbt/s)
In the third trial, we develop a full dynamic bandwidth management system in order to
adjust the human bandwidth to the network bandwidth. When the watcher switches to a 4
screens display, the image resolution becomes smaller so that the image size diminishes
according to the equation (3). As explained previously, the viewer does not notice a
change for over a certain resolution. This is why the low resolution fits perfectly the
small display. Globally the bit rate for the four screen display decreases from 38Mbt/s to
22Mbt/s (Fig. 7). The gain is approximately 43%.
Figure 7 : Evolved dynamic bandwidth management 22Mbt/s – 8Mbt/s
13
This system avoids wasting bandwidth without providing a significant improvement for
the viewer. This method has been implemented in “TrainscopeViewer”, a software
created by the Visual Intelligence Studio group to manage the bandwidth for wireless
network.
5.3 TrainscopeViewer
5.3.1 HTTP request
TrainscopeViewer is an application developed by Visual Intelligence Studio of Carnegie
Mellon University. It manages up to n IP cameras at the same time (tested with 16) and
provides up to 4 different ways to display the cameras (cf. TrainscopeViewer guide). The
software is coded in C++ and uses the Visual .Net compiler. As said previously, an IP
camera is a digital camera and therefore is bidirectional. The video flow can be managed
while sending HTTP requests to the camera. The camera will adapt the flow according to
the requests received dynamically. TrainscopeViewer manage all the cameras available at
the same time, sending requests through the HTTP port. These requests are sent by the
user who is switching between the different views.
http://myserver/cgi/view/videostatus.cgi?status=1,2,3,4?
HTTP/1.0 200 OK\r\n
Content-Type: text/plain\r\n
\r\n
Video 1 = video
Video 2 = no video
Video 3 = no video
Video 4 = video
Figure 8 : HTTP request to check the status of the cameras of the network
http://myserver/mjpg/video.cgi?resolution=320x240&camera=1&compression=25
Figure 9 : HTTP request to modify resolution and compression
http://myserver/cgi/mjpg/video.cgi?fps=5
Figure 10 : HTTP request to modify the frame per second
14
<parameter>=<value>
Values
Description
resolution=<string>
1280x1024, 1280x960,
1280x720, 768x576,
4CIF, 704x576,
704x480, VGA,
640x480, 640x360,
2CIFEXP, 2CIF,
704x288, 704x240,
480x360, CIF, 384x288,
352x288, 352x240,
320x240, 240x180,
QCIF, 192x144,
176x144, 176x120,
160x120 1
Specify the resolution of the returned image.
camera=<string>
1, 2, 3, 4 or quad 1
Applies only to video servers with more than
one video input. Selects the source camera.
compression=<int>
0 - 100 1
Adjusts the compression level of the image.
Higher values correspond to higher
compression, i.e. lower quality and smaller
image size.
colorlevel=<int>2
0 - 100 1
Sets level of color or grey-scale.
0 = grey-scale, 100 = full color.
color=<int>
0, 1
Enables/disables color.
0 = black and white, 1 = color.
clock=<int>
0, 1
Shows/hides the time stamp.
0 = hide, 1 = show.
date=<int>
0, 1
Shows/hides the date.
0 = hide, 1 = show.
rotation=<int>
0, 90, 180, 270 1
Rotates the image clockwise.
textpos=<string>
top, bottom
The position of the string shown in the
image.
overlayimage=<int>2
0, 1
Enable/disable overlay image.
overlaypos=<int>x<int>2
<xoffset>1x<yoffset>1
Set the position of the overlay image.
squarepixel=<int>2
0, 1
Enable/disable square pixel correction.
Applies only to video servers.
Table 4 : HTTP parameters for dynamic management
15
5.3.2 TrainscopeViewer Management diagram
4 screens mode:
Camera 1, 2, 3 and 4
Middle resolution
Middle compression
HTTP requests:
Enable video flow 2, 3 and 4
Camera1,2,3,4 resolution = middle
Camera1,2,3,4 compression = middle
HTTP requests:
Stop video flow 2, 3 and 4
Camera1 resolution = high
Camera1 compression = low
HTTP requests:
Disable video flow 5…16
Camera1,2,3,4 resolution = middle
Camera1,2,3,4 compression = middle
HTTP requests:
Enable video flow 5…16
Camera1…16 = resolution = low
Camera1…16 compression = middle
16 screens mode:
Camera 1, 2, 3 … 16
Low resolution
Middle compression
Full screen mode:
Camera 1
High resolution
Low compression
HTTP requests:
Stop video flow 2, 3 … 16 Camera1
resolution = high
Camera1 = compression = low
HTTP requests:
Enable video flow 5…16
Camera1…16 = resolution = low
Camera1…16 compression = middle
Figure 11 : TrainscopeViewer Management diagram
16
6 Multi-point Accessing
6.1 Network design
The TRAINET network is divided in two parts: land and onboard network. The system is
illustrated in Figure 12 and Figure 13. In this design, the vehicle communicates to the
“Hot-Spot” network on the land through the onboard network. The hot-spots transfer in
real time the data to a router. This one regulates the data flow to the data server. In
addition, dynamic bandwidth management software can automatically monitor the video
stream.
Wireless Bridge
Ex IP: 10.0.0.13
5VDC
SWITCH
7.5VDC
CAMERA 1
Ex IP: 10.0.0.11
CAMERA 2
Ex IP: 10.0.0.12
CAMERA 3
Ex IP: 10.0.0.14
CAMERA N
Ex IP: 10.0.0.N
Figure 12 : Onboard network diagram for TRAINET
Data server Computer
Ex IP: 10.0.0.10
IP Router
Ex IP: 10.0.0.3
Access Point
Ex IP: 10.0.0.1
Power-over-Ethernet
Access Point
Ex IP: 10.0.0.2
Power-over-Ethernet
Access Point N
Ex IP: 10.0.0.N
Power-over-Ethernet
Figure 13 : Land network diagram for TRAINET
17
6.2 Signal Handover
The bandwidth available depends on the network technology and the protocols used to
manage the traffic. However, to understand entirely the bridge management our approach
can not be limited by the consideration of the only link between one wireless bridge and
one access point. The cooperation between the access points generates better results in
terms of range and throughput. The drawbacks of this kind of approach are related to the
protocols used while switching between access points. The previous generation of access
points, they were waiting for the complete loss of the signal before trying to connect to
another access point. The data flow was thereby irregular and could not be applied to
cutting edge wireless applications.
Server
AP
AP
AP
AP
70-110
70-110
70-110
meters
meters
meters
Figure 14 : Network design for access point collaboration
The last generation of access points allows for defining the data rate parameters. So the
threshold data rate can be set to a value T. The access points check all the connections
with the bridge (multipoint) superior to the T value. Then it chooses the best link among
all the links available. The access points act in a “best effort” protocol to enhance the
throughput.
The data rate settings are used to choose the data rates that the access points use for data
transmission. The rates are expressed in megabits per second. The access point always
attempts to transmit at the highest data rate set to Basic, also called Require on the
browser-based interface. If there are obstacles or interference, the access point steps
down to the highest rate that allows data transmission. Data rate (1, 2, 5.5, and 11
megabits per second) can be set to one of three states:
1.
Basic (this is the default state for all data rates)—Allows transmission at this rate for
all packets, both unicast and multicast. At least one of the access point's data rates
must be set to Basic.
2.
Enabled—The access point transmits only unicast packets at this rate; multicast
packets are sent at one of the data rates set to Basic.
18
3.
Disabled—The access point does not transmit data at this rate
In this example, the data rate threshold is equal to 12 Mbt/s (the average bandwidth for 4
cameras is 8 Mbts/s). The configuration settings of the different throughput are described
below.
Network
Throughput
Disabled
1/2/5.5/6 Mbps
9
9 Mbps
9
11 Mbps
9
Basic
12 Mbps
9
18 Mbps
9
Enabled
24 Mbps
9
36 Mbps
9
48 Mbps
9
54 Mbps
9
Table 5 : Configuration of the access point to avoid handovers
The threshold must be chosen according to the average bandwidth used in the network. If
the threshold is too high, the access point will be unable to find a connection with the
bridge. To solve this issue, the access points have to be set closer to each other in order to
have a link with the bridge. However, it decreases the coverage of the wireless network.
The figure 15 shows that the connection between the access points switches later with a
11Mbps threshold.
19
30
Bandwidth Mbps
25
20
Threshold 11 Mbps
Threshold 15 Mbps
15
Poly. (Access Point 1)
Poly. (Access Point 2)
10
5
AP1location: 30 meters
AP2 location: 110 meters
0
0
10
20
30
40 50
60
70
80 90 100 110 120 130
Distance meters
Figure 15 : Signal Handover for 11 and 15 Mbps Threshold
20
6.3 Multiple bridge access points for multiple trains
In a multiple bridge configuration, several transit shuttles establish a communication with
the access points on the land network. In a point-to-multipoint configuration, two or more
non-root bridges associate to a root bridge. According to the wireless expert Cisco, up to
17 non-root bridges can associate to a root bridge, but the non-root bridges must share the
available bandwidth. Therefore, the maximum bandwidth available from the bridge must
be shared between all the access points. Thanks to the signal handover management, the
access point establishes the best link as possible with the bridge (Fig. 16). However, the
point to multipoint range is inferior as the point to point range (Fig. 17).
AP 1
AP 2
AP 3
AP n
Airport Shuttle 1
Airport Shuttle 2
Airport Shuttle n
Figure 16 : Multiple bridge access points for multiple trains
The bridge can be connected on different channels to avoid interferences and increase the
bandwidth. Two pairs of bridges can be set up to add redundancy or load balancing to the
bridge link. The bridges must use non-adjacent, non-overlapping radio channels to
prevent interference, and they must use Spanning Tree Protocol (STP) to prevent bridge
loops. The point to multipoint protocol “Frame relay” is used for the access points (Cisco
Aironet 1300 series).Frame Relay is a networking protocol that works at the bottom two
levels of the OSI reference model: the physical and data link layers. It is an example of
packet-switching technology, which enables end stations to dynamically share network
resources. It provides global addressing, Virtual Circuit status messages and multicasting.
Frame Relay devices fall into the following two general categories:
• Data terminal equipment (DTEs), which include terminals, personal computers,
routers, and bridges
21
• Data circuit-terminating equipment (DCEs), which transmit the data through the
network and are often carrier-owned devices (although, increasingly, enterprises are
buying their own DCEs and implementing them in their networks)
Frame Relay networks transfer data using one of the following two connection types:
• Switched virtual circuits (SVCs), which are temporary connections that are created
for each data transfer and then are terminated when the data transfer is complete (not a
widely used connection)
•
Permanent virtual circuits (PVCs), which are permanent connections
The DLCI is a value assigned to each virtual circuit and DTE device connection point in
the Frame Relay WAN. Two different connections can be assigned the same value within
the same Frame Relay WAN—one on each side of the virtual connection.
22
7 Distance issues
For long network over 300 meters, the handover management is an important factor.
However, the antennas are also crucial components to provide the best range and
throughput. Two different high power antennas can be used: omnidirectionnal and
directional antennas. In this investigation, only omnidirectionnal antennas for 802.11g
are used.
Network traffic
in Megabit per second
30
25
20
Antenna 3dbi
15
Antenna 10dbi
10
5
0
0
10
20
30
40
50
60
70
80
Distance from access point in meters
Figure 17 : Network traffic vs. distance
The graphic above (Fig.16) compares the throughput versus the distance in an outdoor
environment for two antennas. The first antenna (3dbi) is a standard antenna and the
throughput decreases quickly after 20 meters. The 10dbi omnidirectionnal antenna has a
better range and the throughput decreases only after 50 meters. The maximum throughput
observed is 26 Mbps (Fig 17) at a distance of 10 meters from the access point.
Figure 18 : Maximum bandwidth for outdoor conditions 26 Mbps
23
The throughput is directly linked to the linking quality of the bridge. It depends on the
distance, the weather and the ‘line of sight’. The figure 18 represents the linking quality
in an outdoor environment with the “line of sight”.
Figure 19 : Linking Quality vs. Distance from access points
There is a difference between the values claimed by the manufacturer (Table 6) and the
values observed during our tests. The antenna range observed is represented Figure 19.
The area within 45 meters from the access point has a throughput over 22 Mbps. Only
this area of interest for our application.
Range
802.11g: Outdoor
110 ft (34m) @ 54 Mbps
200 ft (61 m)@ 48 Mbps
225 ft (69 m) @ 36 Mbps
325 ft (99 m) @ 24 Mbps
400 ft (122 m) @ 18 Mbps
475 ft (145 m) @ 12 Mbps
490 ft (149 m) @ 11 Mbps
550 ft (168 m) @ 9 Mbps
650 ft (198 m) @ 6 Mbps
660 ft (201 m) @ 5.5 Mbps
690 ft (210 m) @ 2 Mbps
700 ft (213 m) @ 1 Mbps
Table 6 : Network range for aironet
1200 Access Point (Cisco)
10 Mbps
90 meters
22 Mbps
45 meters
Figure 20 : Coverage of the omni directional antenna 10 dbi
24
As with all radio systems, interference is always a problem. If we are listening to an AM
radio and we hear static, this is interference. The same thing applies to WiFi systems,
however not to such a large degree. Things that cause interference with WiFi systems are
Microwave ovens, certain lighting systems, other 802.11 access points or systems,
microwave transmitters, even high speed processors for computers can cause interference
for 802.11 systems. All these problems must be isolated before we can expect any
significant range out of our system.
The use of directional antenna depends on the network’s design. It can increase
significantly the range and the throughput for specific designs and improve the security
of the network. At the beginning and at the end of the track, directional antenna should be
set up to avoid transmitting the signal outside of the track where it could be intercepted.
25
8 Video Quality
Image quality is clearly one of the most important features of any camera, if not the most
important. This is particularly so in surveillance and monitoring applications, where lives
and property may be at stake. By developing image processing chips and sophisticated
algorithms tailored for network camera applications, image quality has been improved to
a degree never before seen at lower cost levels. As digital technology becomes
commonplace and replaces analog solutions, there will be further advances in areas such
as high resolution and advanced video compression, but success will ultimately depend
on how well the initial information is captured and handled.
Superior image quality enables the user to:
•
more closely follow details and changes in images, making for better and faster
decisions concerning the safety of people and property
•
With greater accuracy use automated analysis and alarm tools, such as face
recognition, with fewer false positives.
Unlike traditional analog cameras, digital network cameras are equipped with the
processing power not only to capture and present images, but also to digitally manage
and compress them for network transport. Image quality can vary considerably and is
dependent on the choice of optics and image sensor, the available processing power and
the level of sophistication of the algorithms in the processing chip.
¾ Image sensor
There are two types: CCD (Charged Coupled Device) and CMOS (Complementory metal
oxide semiconductor) CCD sensors are produced using a technology developed
specifically for the camera industry, while CMOS sensors are produced by the same
technology used for the chips used in computers. Today's high quality cameras mostly
use CCD sensors. Although recent advances in CMOS sensors are closing the gap, they
are still not suitable for cameras where the highest possible image quality is required.
¾ Image resolution
Higher resolution means more detail, and as cameras now deploy megapixel sensors that
make it possible to capture even more detail, analog CCTV cameras--which are bound to
resolutions used in TV standards--are being surpassed.
26
Vehicle 1 – Camera 1
Vehicle 1 – Camera 2
Vehicle 2 – Camera 3
Vehicle 2 – Camera 4
Figure 21 : Image quality with backlight compensation
¾ Backlight compensation
While a camera's automatic exposure control tries to get the lightness of an image to
appear as the human eye would see a scene, it can be easily fooled. Think of the case
where a person walks into a fairly dark room with a flashlight in her hand and directs this
flashlight to the camera. Although the light source is quite small, it makes the camera
believe the scene has become brighter and the camera's exposure control automatically
adjusts to it, resulting in a darker image. To avoid this, a mechanism called backlight
compensation is introduced. It strives to ignore small areas of high illumination, just as if
they were not present at all. With backlight compensation, the image from the example
above would have the same exposure regardless of whether the flashlight was present or
not. The resulting image enables the person to be visually seen and identified. Without
backlight compensation, the image would be too dark, and identification would be
impossible.
¾ Ability to correctly capture moving objects
In addition to good light sensitivity, another key feature to look for is progressive scan.
That the camera has progressive scan means that images do not suffer from the "saw"
effect that hampers interlaced video technologies The interlace mode is used in TVs and
traditional analog CCTV cameras in order to enhance the image frequency in moving
images. The "saw" effect becomes apparent when the picture is frozen. Soon, network
cameras and IP-Surveillance technology will deliver superior image quality by means of
mega-pixel resolution. Analog cameras are limited by the 0.4 Megapixel resolution of
NTSC/PAL standards.
27
9 Scalability
9.1 Storage
When the basis for the IP networking architecture was developed in the 1960s and '70s,
the ability to provide redundancy was the top requirement. In the same way today,
transmission links, application servers, storage and switches can all have parallel layers
of services and alternative routes of communications.
Storage can be consolidated to secure off-site locations, and servers can use redundant
power supplies, hot-swap RAID disks, error-correcting memory and dual network cards.
This is all up to the network designer, and although a small network will not deploy all of
the possible safety measures, choosing high-quality IT components in the network is in
any case likely to be a more reliable solution than CCTV with VCRs or black box DVRs.
And don't forget, by using standard server and network equipment, replacing faulty
hardware takes much less time and is less costly than with proprietary DVR solutions.
Scalability steps: A DVR system is usually supplied with 4, 9 or 16 camera inputs,
therefore becomes scalable in steps of 4, 9 or 16. If a system includes 15 cameras, this is
not an issue, but it becomes a problem if 17 cameras are needed. Adding one single
camera would generate the need for an additional DVR. Network video systems are far
more flexible, and can be scaled in steps of one camera at a time.
Number of cameras per recorder: In a network video system, a PC server records and
manages the video. The PC server can be selected according to the performance needed.
Performance is often specified as number of frames per second, total for the system. If 30
fps is needed for each camera, one server may only record 25 cameras. If 2 fps is
sufficient, 300 cameras can be managed by one server. This means that the performance
of the system is used efficiently and can be optimized.
Size of system: For larger installations, a network video system is easy to scale. When
higher recording frame rates or longer recoding times are needed, more processing and/or
memory capacity can be added to the PC server managing the video. Even more simply,
another PC server can be added, located either at a central location, or at remote
locations.
9.2
9.2.1
Network Hardware
Digital Cameras scalability
The system was tested with up to 5 digital cameras, but the number of cameras in a
vehicle has no limit (contrary to an analog system). It only depends on the number of
ports on the switch (5-80). The limiting factor is the bridge connection while some peaks
up 26 Mbps were observed, only the average bandwidth of 22 Mbps will be considered in
this study.
28
254 cameras
Bandwidth Mbps
20
240x180
15
320x240
10
640x480
5
198
187
176
165
154
143
132
121
110
99
88
77
66
55
44
33
22
11
0
0
Number of cameras
Figure 22 : Bandwidth vs. Number of camera per vehicle at 1 fps for different resolution
A vehicle is equipped with several cameras (compression 30%, 1 fps) and tested with
three different resolutions (240x180, 320x240, 640x480). With one bridge, the bandwidth
can be shared up to 4 cameras in 640x480, up to 8 in 320x240 and up to 14 in 240x180.
4 cameras
25
Bandwidth Mbps
20
240x180
15
320x240
10
640x480
5
0
0
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
Number of cameras
Figure 23 : Bandwidth vs. Number of camera per vehicle at 30 fps for different resolution
A vehicle is equipped with several cameras (compression 30%, 30 fps) and tested with
three different resolutions (240x180, 320x240, 640x480). With one bridge, the bandwidth
can be shared up to 4 cameras in 640x480, up to 8 in 320x240 and up to 14 in 240x180.
29
4 cameras
25
Bandwidth Mbps
20
240x180
15
320x240
10
640x480
5
0
0
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
Num ber of cam eras
Figure 24 : Bandwidth vs. Number of camera per vehicle at 15 fps for different resolution
A vehicle is equipped with several cameras (compression 30%, 15 fps) and tested with
three different resolutions (240x180, 320x240, 640x480). With one bridge, the bandwidth
can be shared up to 8 cameras for a resolution of 640x480.
4 cameras
Bandwidth Mbps
25
20
compression
50%
compression
30%
compression
10%
15
10
5
198
180
162
144
126
108
90
72
54
36
18
0
0
Num ber of cam eras
Figure 25 : Bandwidth vs. Number of camera per vehicle at 1 fps for different compressions
A vehicle is equipped with several cameras (resolution 320x240, 30 fps) and tested with
three different compressions (10%, 30%, 50%). With one bridge, the bandwidth can be
shared up to 6 cameras in 640x480, up to 9 in 320x240 and up to 11 in 240x180.
30
4 cameras
B an d w id th M b p s
25
20
compression
50%
compression
30%
compression
10%
15
10
5
0
0
1
2
3
4
5
6
7
8
9
10
11
Number of cameras
Figure 26 : Bandwidth vs. Number of camera per vehicle at 30 fps for different compressions
A vehicle is equipped with several cameras (resolution 320x240, 30 fps) and tested with
three different compressions (10%, 30%, 50%). With one bridge, the bandwidth can be
shared up to 6 cameras in 640x480, up to 9 in 320x240 and up to 11 in 240x180.
4 cameras
Bandwidth Mbps
25
20
compression
50%
compression
30%
compression
10%
15
10
5
0
0
1
2
3
4
5
6
7
8
9 10 11 12 13 14
Number of cameras
Figure 27 : Bandwidth vs. Number of camera per vehicle at 15 fps for different compressions
A vehicle is equipped with several cameras (resolution 320x240, 15 fps) and tested with
three different compressions (10%, 30%, 50%). With one bridge, the bandwidth can be
shared up to 12 cameras in 640x480.
31
9.2.2 Increasing the length of the track
<280 meters
<540 meters
<810 meters
<1080 meters
<1350 meters
<1620 meters
= 1mile
1
2
3
4
5
Ethernet cable
(90 meters)
3
6
9
12
15
Optical Switch
(number of ports)
No optical
7
10
13
16
6
18
19
Access Point*
Switch
3
6
9
12
15
18
Table 7 : Network configuration for several network range*
*The number of access points is available for an omnidirectional antenna 3dbi. The use
of different antennas (directional, 10 dbi) would reduce this number.
The table 7 describes the scalability of the system for different lengths of the track. For a
length under 280 meters, Ethernet cable can be used instead of optical fiber. For longer
track, the optical switch can interconnect the different access points. Moreover, the
optical fiber has a bandwidth available equal to several Giga bits per seconds.
9.2.3 Increasing the number of vehicles
70
Bandwidth Mbps
60
50
40
30
4 cameras
6 cameras
22 Mbps
20
10
0
1
2
3
4
5
6
Number of vehicle
Figure 28 : Bandwidth vs. Number of vehicle at 30 fps 320x240 30%
Theoretically, each vehicle is equipped with four or six cameras (resolution 320x240, 15
fps, compression 30%). In a point to multipoint mode, the vehicles share the available
bandwidth on the same channel. With vehicles equipped with six cameras, the system can
manage two vehicles and up to three vehicles equipped with four cameras. This is the
maximum number connection per access point. More vehicles can be added but they do
not have to connect the same access point at the same time.
32
Number of
bridges
3
2
1
6 cameras
4 cameras
0
1
10
20
30
40
50
60
70
80
90
Num ber of vehicles
100
4 cameras
6 cameras
Figure 29 : Number of bridges vs. number of vehicles (4 cameras or 6 cameras) at 1 fps
Number of
bridges
3
2
1
6 cameras
4 cameras
0
1
2
3
4
Number of vehicles
5
6
4 cameras
6 cameras
Figure 30 : Number of bridges vs. number of vehicles (4 cameras or 6 cameras) at 30 fps
A way to tackle this problem is to add several bridges. Two pairs of bridges can be set up
to add redundancy or load balancing to the bridge link. The bridges must use nonadjacent, non-overlapping radio channels to prevent interference, and they must use
Spanning Tree Protocol (STP) to prevent bridge loops. The figure 26 shows the
theoretical results for vehicle equipped with four or six cameras (resolution 320x240, 30
fps, compression 30%). The figure 27shows the theoretical results for vehicle equipped
with four or six cameras (resolution 320x240, 15 fps and compression 30%).
33
Number of
bridges
3
2
1
6 cameras
4 cameras
0
1
2
3
4
5
6
7
8
Number of vehicles
4 cameras
6 cameras
Figure 31 : Number of bridges vs. number of vehicles (4 cameras or 6 cameras) at 15 fps
9.3
Pittsburgh Technology Center test for throughput
Several tests have been realized in the Pittsburgh Technology Center (PTC) for
measuring the throughput. The goal was to reproduce the results obtained in the indoor
laboratory. The parameters such as weather, line of sight and vehicle speed were
unknown factors for the network throughput. The network is configured as described
Figure 32 and Figure 33.
Wireless Bridge
IP: 10.0.0.13
SWITCH
CAMERA 1
IP: 10.0.0.11
CAMERA 2
IP: 10.0.0.12
CAMERA 3
IP: 10.0.0.14
CAMERA 4
IP: 10.0.0.15
Figure 32 : Onboard network diagram for PTC TEST
34
Data server Computer
IP: 10.0.0.13
IP Router
Ex IP: 10.0.0.3
Access Point
Ex IP: 10.0.0.1
Power-over-Ethernet
Access Point
Ex IP: 10.0.0.2
Power-over-Ethernet
Figure 33 : Land network diagram for PTC TEST
The test road is 300 meters long (Fig. 36 yellow line) and its configuration is close to the
design of the test track in Bombardier Transportation. Two access points are set up along
the track (Fig. 36 white points) to cover the all distance. The access points have a default
configuration and all the different network throughput are enabled.
35
Figure 34 : Camera set up 1
Figure 35 : Camera set up 2
Network
Throughput
DiCamera set
uosabled
1/2/5.5/6/9
Mbps
9
Enabled
Basic
11 Mbps
9
12 Mbps
9
18 Mbps
9
24 Mbps
9
36 Mbps
9
48 Mbps
9
54 Mbps
9
Table 8 : Configuration of the access points (Access Point 1 & 2)
36
Figure 36 : Satellite view Pittsburgh Technology Center test
Yellow line: test road (≈300 meters)
White Points: Access points
The resulting coverage is about 220 meters. However, two zones must be differentiated:
the middle zone and the border zone. The handover is smooth between the access points
and the throughput is about 24 Mbps (Table 9). The configuration is really efficient in
the middle zone and the results match the criteria of the project in terms of resolution,
number of cameras, distance and throughput. The throughput used for the video is under
the half of the bandwidth available (43%). The bandwidth available is shifting due to the
handover, this issue is solved in the next chapter. The connection in the border zone is
poor and some disconnections happen at the very end of the track. The frame rate is slow
(10 fps) but the quality of the picture is still good due to the MJPEG compression. Each
frame has the quality of a JPEG image.
Frame per second
Distance from access
point
Resolution
Compression
Bandwidth available
Bandwidth used
Delay
Disconnection
Middle Zone
(between access
points)
30
40
Border Zone
(between access point
and end of the track)
10
50
320x480
10%
24 Mbps
10 Mbps (43%)
No
No
320x480
10%
14 Mbps
5 Mbps (22%)
Yes
Yes at the end
Table 9 : Network Measures for the Pittsburgh Technology Center
37
Number of cameras
4
Number of access points
2
Number of bridge
1
Bandwidth
14 -24 Mbps
Resolution
320x480
Antenna
10dbi
Compression
10%
Distance covered with network connection 250 meters
Distance covered with throughput >22 Mps 190 meters
Vehicle speed
40-50 mph
Table 10 : Results for the Pittsburgh Technology Center
9.4 Pittsburgh Technology Center Test for Handover
The onboard network remains the same as for the test for the throughput. An access point
is added to the land network in order to test the reliability of the handover (Fig. 37).
Three access points are set up along the track at the distance 60, 140 and 220 meters (Fig.
38). The configuration of the access points (Fig. 38) were changed to maximize the
throughput when switching between access points. The threshold T is set gradually to 11,
12, 18 Mbps.
Data server Computer
IP: 10.0.0.13
IP Router
Ex IP: 10.0.0.3
Access Point
Ex IP: 10.0.0.1
Power-over-Ethernet
Access Point
Ex IP: 10.0.0.2
Power-over-Ethernet
Access Point
Ex IP: 10.0.0.4
Power-over-Ethernet
Figure 37: Land network diagram for PTC TEST
38
Network Throughput
Disabled
1/2/5.5/6/9 Mbps
9
Basic
11 Mbps
9
12 Mbps
9
18 Mbps
9
Enabled
24 Mbps
9
36 Mbps
9
48 Mbps
9
54 Mbps
9
Table 11 : Configuration of the access points for a threshold equal to 11Mbps
Figure 38 : Satellite view Pittsburgh Technology Center test
Yellow line: test road (≈300 meters)
White Points: Access points
The throughput is measured along the track for the different thresholds (Fig. 39) and the
disconnection are indicated with bandwidth available equal to zero. Therefore, a
threshold at 11 Mbps has a better coverage (320 meters but the average bandwidth is 1720 Mbps). For the 12 Mbps threshold, the average bandwidth is 20-22 Mbps for coverage
of 280 meters whereas for the 18 Mbps, the coverage is only 200 meters but with an
average bandwidth of 21-23 Mbps. The investigators observed a disconnection at the
distance 180 meters of the track. This drawback is due to the threshold which is too high.
Neither of the access point can find a link with the bridge with a 18 Mbps threshold. That
is the reason why a threshold of 12 Mbps will be used in the tests. It is the best
compromise between coverage, throughput and handover.
39
Handover management for different thresholds
25
11 Mbps
15
12 Mbps
18 Mbps
10
5
300
280
260
240
220
200
180
160
140
120
100
80
60
40
20
0
0
Bandwidth (Mbps)
20
Test track distance (meters)
Figure 39 : Bandwidth vs. distance (thresholds 11 Mbps, 12 Mbps and 18 Mbps)
Number of cameras
4
Number of access points
3
Number of bridge
1
Bandwidth
20-22 Mbps
Threshold for handover
12 Mbps
Resolution
320x480
Compression
10%
Distance covered with throughput >22 Mps 280meters
Vehicle speed
40-50 mph
Table 12 : Results for the Pittsburgh Technology Center – Handover test
40
9.5 Pittsburgh Technology Center test for multiple bridges
Three access points are set up along the track at the distance 60, 140 and 220 meters
(Fig.40 and Fig.41). The access points are set up in point to multipoint mode. The land
network remains the same as the test for the handover is concerned. There are two mobile
onboard networks. Each one includes two IP cameras fastened to the car frames. Two
scenarios are tested:
1.The cars follow each other in order to share the same connection (same access point).
2.The cars are coming from two different directions and intercept in the middle of the test
track.
Wireless Bridge
IP: 10.0.0.13
Wireless Bridge
IP: 10.0.0.19
SWITCH
SWITCH
CAMERA 1
IP: 10.0.0.11
CAMERA 2
IP: 10.0.0.12
Figure 40 : Onboard network for vehicle 1
CAMERA 3
IP: 10.0.0.13
CAMERA 4
IP: 10.0.0.14
Figure 41 : Onboard network for vehicle 2
9.5.1 Scenario 1: Cars following each other
In this scenario, each access point has to handle two bridges connection. The video
quality is the same as the point to multipoint connection (Image Quality 320x480,
Compression 10%). The vehicles follow each other at a speed of 20-25 mph and use the
road track of the Pittsburgh Technology Center (Fig. 42).
41
Vehicle 1 – Camera 1
Vehicle 1 – Camera 2
Vehicle 2 – Camera 3
Vehicle 2 – Camera 4
Figure 42 : Camera screenshots for multiple bridges scenario 1
AP 1
AP 2
AP 3
Vehicle 1
Vehicle 1
Vehicle 2
Vehicle 2
30 Frames per second
Image Quality 320x480
Compression 10%
10 Frames per second
Image Quality 320x480
Compression 10%
Figure 43 : Multiple bridges with two vehicle following each other
The frame rate is 30 frames per second between two access points but decreases quickly
when the two bridges are connected to the same access point (Fig. 43). The throughput
received by the network is 10 Mbps and is the same as in a point to point mode. To
maintain, a frame rate equal to 30 frames per second, the vehicle must always be between
two access points. When the two bridges are connected to the same access point, that
42
point must handle all the communication and it introduces a delay in the video flow
processing.
Number of cameras
Number of access points
Number of bridge
Antenna
Bandwidth
Threshold for handover
Resolution
Compression
Distance covered with throughput >22 Mps
Vehicle speed
Frame per seconds
2x2
3
2
10 dbi
20-22 Mbps
12 Mbps
320x480
10%
230meters
40-50 mph
30 but only 10
at the end of the track
Table 13 : Results for the Pittsburgh Technology Center – multiple bridges test1
9.5.2 Scenario 2: Cars intercepting
The video quality is the same as the point to multipoint connection (Image Quality
320x480, Compression 10%). The vehicles intercept each other at a speed of 30-40 mph
and use the road track of the Pittsburgh Technology Center (Fig. 45). The cars start from
the two extremity of the track and drive in an opposite direction to intercept in the middle
of the track.
Vehicle 1 – Camera 1
Vehicle 1 – Camera 2
Vehicle 2 – Camera 3
Vehicle 2 – Camera 4
Figure 44 : Camera screenshots for multiple bridges scenario 1
43
AP 1
AP 2
AP 3
Time T=0s
Vehicle 2
Vehicle 1
30 Frames per second
Image Quality 320x480
Compression 10%
AP 1
AP 2
AP 3
Vehicle 2
Time T=30s
Vehicle 1
30 Frames per second
Image Quality 320x480
Compression 10%
Figure 45 : Multiple bridges with two vehicles intercepting
This test is close to the real application with two airport shuttles making round-trips
between the airport and the planes. At the time T=0s, the connection is point to
multipoint and the bandwidth available for each connection is 22 Mbps. In this test, 30
frames per seconds are observed from T=0s to T=25s. When the vehicles intercept in the
middle of the track (T=25s to T=30s), a brief slowdown (<2s) is observed. In the last test,
30 frames per seconds are observed in the point to multipoint mode. This value confirms
the results of the previous test with two vehicles following each other between two access
points.
Number of cameras
2x2
Number of access points
3
Number of bridge
2
Antenna
10 dbi
Bandwidth
20-22 Mbps
Threshold for handover
12 Mbps
Resolution
320x480
Compression
10%
Distance covered with throughput >22 Mps
280meters
Vehicle speed
30-40 mph
Frame per seconds
30 but only 10
at the end of the track
Table 14 : Results for the Pittsburgh Technology Center – multiple bridges test2
44
9.6
Bombardier Transportation Test
9.6.1 System Configuration
The final test was realized on the Bombardier Transportation’s test track in Pittsburgh.
The design of the track is presented below (Fig. 46). The track is 300 meters long. For
this distance, three access points are set up along the track (AP1:55 meters, AP2: 145
meters, AP3: 235 meters).
Figure 46 : Bombardier Transportation test track satellite view
Yellow line: test road (≈300 meters)
White Points: Access points
Figure 47 : Shuttle vehicle on the test track
45
9.6.2 Network settings
The network is configured to switch when the connection bandwidth is under 12Mbit/s.
The cameras located in the moving vehicle are managed dynamically by the software
“TrainscopeViewer”.
Wireless Bridge
Ex IP: 10.0.0.13
SWITCH
CAMERA 1
IP: 10.0.0.11
CAMERA 2
IP: 10.0.0.12
CAMERA 3
IP: 10.0.0.14
CAMERA 4
IP: 10.0.0.15
Figure 48 : Onboard network diagram for TRAINET
Data server Computer
IP: 10.0.0.10
IP Router
IP: 10.0.0.3
Access Point
IP: 10.0.0.1
Access Point
IP: 10.0.0.2
Access Point 3
IP: 10.0.0.4
Figure 49 : Land network diagram for TRAINET
9.6.3 Field results
With the help of a camera recorder, the video flow is recorded. The flow is smooth and
30 frames per seconds are observed. However, some packets are lost during the wireless
transfer. It results some slowdown while transferring the video. This drawback can be
avoided while fastening the access points to a post. It will avoid most of the reflection
with the ground. The image quality is good and we can clearly identify building and
people moving inside the vehicle.
46
Figure 50 : Image recorded through the wireless network
The frames are recorded through the wireless network with a resolution of 320x240
pixels and a compression of 50%. The format for the video is MJPEG, therefore the still
images have the quality of JPEG. Some recorders can be added onboard to record the
video flow for saving the video in high definition.
Due to the dynamic bandwidth management only 10Mbps are used the worst case. The
average bandwidth used is 8 Mbt/s and is far below the capacity of the wireless link (2022Mbts). This low bandwidth is linked to the low complexity of the scene (see the last
appendix of the published paper). Most of the cameras were aimed outside and half of the
pictures represented the sky. The movement of the people in the vehicle will increase the
complexity of the scene and by consequence the bandwidth used (Annexe 1D-1E).
Figure 51 : Network traffic for Bombardier Test track (average 8 Mbps) (100% 100Mbps)
47
Number of cameras
Number of access points
Number of bridge
Antenna
Bandwidth
Threshold for handover
Resolution
Compression
Distance covered with throughput >22
Mps
Vehicle speed
Frame per seconds
4
3
1
10 dbi
20-22 Mbps
11 Mbps
320x480
10%
280meters
30-40 mph
20-30 fps
Table 15 : Results for Bombardier Transportation Test
48
Access Point 1
Access Point 2
Access Point 3
Stop
Start
280 meters
Fig. 45: Bombardier test track design
49
Glossary of acronyms
16-QAM — 16-Phase Quadrature Amplitude Modulation
64-QAM — 64-Phase Quadrature Amplitude Modulation
ADC — Analog-to-Digital Converter
AES — Advanced Encryption Standard
AHB — Advanced High-performance Bus
APB — Advanced Peripheral Bus
ARM-CPU — A Central Processing Unit based
on Intellectual Property Licensed from ARM
ATM — Asynchronous Transfer Mode
BPSK — Binary Phase-Shift Keying
BWA — Broadband Wireless Access
CPE — Customer Premises Equipment
CRC — Cyclic Redundancy Code
DAC — Digital-to-Analog Converter
DES — Data Encryption Standard
DMA — Direct Memory Access
DSL — Digital Subscriber Line
E1 — a dedicated phone connection
supporting data rates of 2 Mb/s
ECC — Error Correction Code
FDD — Frequency Division Duplexing
FFT — Fast Fourier Transform
IEEE — Institute of Electrical and Electronics Engineers
IP — Internet Protocol
LOS — Line-of-Sight
MAC — Media Access Control
MMDS — Multichannel Multipoint Distribution Service
NLOS — Non-Line-of-Sight
OFDM — Orthogonal Frequency Division Multiplexing
PDA — Personal Digital Assistant
PHY — Physical Layer
PMP — Point-to-Multipoint
QPSK — Quadrature Phase-Shift Keying
RF — Radio Frequency
RISC — Reduced Instruction Set Computer
RX — Receiver
SoC — System on Chip
T1 — a dedicated phone connection supporting
data rates of 1.544 Mb/s
TDD — Time Division Duplexing
TX — Transmitter
50
Hardware list SUMMARY
Component
Task
Power
Linksys
WET54Gv2
Wireless
bridge
Netgear 5
Ports
Switch
10W
5V-DC / 2A
7.5W
7.5V-DC /
1A
Axis 210
Digital
Camera
10W
Vendor
App.
price
Datasheet
Linksys
$110 US
1-A
Netgear
$50 US
1-B
Axis
$450 US
1-C
Table.1. Hardware list for the TRAINET’s onboard network
Linksys WET54Gv2
Netgear 5 Ports
Component
Task
Power
IBM Thinkpad
Data server
Computer
Cisco Catalyst
3500 series XL
Router
Cisco Aironet
1200 Series
Ethernet Cable
90 meters
Wireless
Access Points
Link
components
16V-DC /
4.5A
7.5W
7.5V-DC /
1A
48V-DC /
380mA
-
Axis 210
Vendor
App.
price
Datasheet
IBM
$1410 US
-
Cisco
$100 US
1-D
Cisco
$450 US
1-E
Pittsburgh
Wires
$50 US
-
Table.1. Hardware list for the TRAINET’s onboard network
51
Cisco Aironet 1200 Series
Cisco Catalyst 3500 series
XL
Ethernet Cable
52
Annexe 1A - Linksys WET54Gv2 Datasheet
Features
Product Description - Linksys Wireless-G Ethernet Bridge WET54G - wireless access point
Device Type - Wireless access point
Enclosure Type – External
Dimensions (WxDxH) - 5 in x 4.2 in x 1.1 in
Weight - 0.4 lbs
Data Link Protocol - Ethernet, Fast Ethernet, IEEE 802.11b, IEEE 802.11g
Network / Transport Protocol - TCP/IP
Features - MDI/MDI-X switch
OS Required - Microsoft Windows 98 Second Edition / Windows ME, Microsoft Windows 2000 / XP
Technical Specs
Device Type - Wireless access point
Width - 5 in
Depth - 4.2 in
Height - 1.1 in
Weight - 0.4 lbs
Form Factor - External
Connectivity Technology - Wireless
Data Transfer Rate - 54 Mbps
Line Coding Format - DBPSK, DQPSK, CCK, OFDM
Data Link Protocol - Ethernet, Fast Ethernet, IEEE 802.11b, IEEE 802.11g
Network / Transport Protocol - TCP/IP
Frequency Band - 2.4 GHz
Status Indicators - Port status, power
Features - MDI/MDI-X switch
Encryption Algorithm - 128-bit WEP, 64-bit WEP
Compliant Standards - IEEE 802.3, IEEE 802.3U, IEEE 802.11b, IEEE 802.11g
Antenna - Detachable
Antenna Qty - 1
Interfaces - 1 x network - Ethernet 10Base-T/100Base-TX - RJ-45 1 x network - Radio-Ethernet
53
Cables Included - 1 x network cable
Compliant Standards - FCC Class B certified
Power Device - Power adapter - external
Software Included - Drivers & Utilities
OS Required - Microsoft Windows 98 Second Edition / Windows ME, Microsoft Windows 2000 / XP
Min Operating Temperature - 32 F
Max Operating Temperature - 104 F
Humidity Range Operating - 10 - 90%
54
Annexe 1B – NETGEAR SWITCH 5 PORTS
The FS100 series Fast Ethernet switches brings the 100 Mbps switching technology in a
compact form factor to the small office marketplace at an aggressive price. These
switches segment networks for improved performance, enabling the most demanding
multimedia and imaging applications. Since each port is auto-speed-sensing, individual
hubs or directly attached servers can easily be upgraded from 10 to 100 Mbps.
Each port can automatically negotiate the network speed and duplex mode with the
remote end, taking the burden of configuration off the user. At the higher speed running
in full duplex mode, each port supports up to 200 Mbps of information throughput. With
it’s speed and it’s reliability, the FS100 series is the best choice for your networking
needs.
Main Specs:
Product Description:
Device Type:
Form Factor:
Dimensions (WxDxH):
Weight:
Ports Qty:
Data Transfer Rate:
Data Link Protocol:
Communication Mode:
Features:
Compliant Standards:
Manufacturer Warranty:
NETGEAR FS105 - switch - 5 ports
Switch
External
6.2 in x 4.1 in x 1.1 in
1.3 lbs
5 x Ethernet 10Base-T, Ethernet 100Base-TX
100 Mbps
Ethernet, Fast Ethernet
Half-duplex, full-duplex
Full duplex capability, uplink, MDI/MDI-X
switch
IEEE 802.3U-LAN, IEEE 802.3i-LAN, IEEE
802.3x
5 years warranty
55
Annexe 1C – AXIS 210 DIGITAL CAMERA
General
• Motion JPEG or MPEG-4 based network
camera with built-in web server
• Superior quality CCD image sensor with
progressive scan
Security
• Multi user level password protection for
restricted camera access
• IP address filtering
System Requirements
The following specification applies to the use of
browser-based viewing. When using application
software, please refer to the specifications
provided with the software.
• Supported operating systems:
Windows XP, 2000, NT4.0, ME and 98, Linux
and Mac OS X
• Supported Web Browsers:
Windows - Internet Explorer 5.x or later and
Mozilla 1.4* or later
Linux - Mozilla 1.4* or later
Mac OS X - Mozilla 1.4* or later and Netscape
7.1* or later
• Hardware:
Meeting the specification for selected
operating system and browser
Hardware & System
• AXIS ETRAX 100LX 32bit RISC CPU
• 16 MB RAM, 4 MB Flash
• AXIS ARTPEC-2 video compression chip with
8 MB RAM
• Linux 2.4 kernel
• Watch Dog functionality
Connections
• Network: 10Base-T/100Base-TX Ethernet
networks (RJ-45)
• I/O: 1 alarm input + 1 output (terminal block)
• Power: 9 VDC / 9 W - external Power Supply,
included
• Alternative input voltage 7-20V DC, min 4W
Camera
• Lens F2.0 4 mm glass lens, C/CS mount
• Focus range 0.5 mm to infinity
• Sony Super HAD progressive scan CCD 1/4”
RGB
• Color & Black/White
• Frame rate: up to 30 frames per second in all
resolutions
• Video compression: Motion JPEG and MPEG-4
(Part 2, Advanced Simple Profile at level 5)
• Supports various resolutions and video quality
settings
• Max resolution: 640 x 480 VGA in 30 frames
per second
• Illumination: 3-10 000 Lux
• Rotation: 90, 180, 270 degrees
• Image scaling
The file size of a JPEG image depends on
resolution, compression level and image
content. Below is a table with average file size
in KByte, derived from real life tests
Functions
• Built-in Video Motion Detection
• Scheduled and triggered event functionality
with alarm notification via e-mail, TCP, HTTP
and upload of images via e-mail, FTP & HTTP
• Pre/post alarm buffer of up to 20 seconds
of 320x240 resolution video at 4 frames per
second
• A mask, custom logo or overlay may be
inserted in the video stream
• Flash memory provide ability to upload
embedded applications
• Up to 20 simultaneous users
• UPnP
• AXIS Dynamic DNS service
Firmware updates
• Flash memory allows firmware updates
over the network using HTTP or FTP over
TCP/IP. Firmware upgrades are available from
www.axis.com
56
Annexe 1D – Cisco Catalyst 3500 series XL
Description
Specification
Performance
10.8 Gbps switching fabric
8.8 million packets-per-second forwarding rate (Catalyst 3548 XL)
7.5 million packets-per-second forwarding rate (Catalyst 3508G XL)
6.5 million packets-per-second forwarding rate (Catalyst 3524 XL)
4.8 million packets-per-second forwarding rate (Catalyst 3512 XL)
(All forwarding rates for 64-byte packets)
5.4 Gbps max forwarding bandwidth
4 MB shared-memory architecture shared by all ports
Packet forwarding rate for 64-byte packets: 14,880 PPS to 10 Mbps
ports 148,800 PPS to 100BaseT ports 1,488,000 PPS to 1000BaseX
ports
8192 Media Access Control (MAC) addresses
8 MB DRAM and 4 MB Flash memory onboard
Management
IEEE 802.3x full duplex on 10BaseT, 100BaseTX, and 1000BaseX
ports
IEEE 802.1D Spanning-Tree Protocol
IEEE 802.1Q VLAN
IEEE 802.3z 1000BaseX specification
1000BaseX (GBIC)
1000BaseSX
1000BaseLX/LH
IEEE 802.3u 100BaseTX specification
IEEE 802.3 10BaseT specification
Y2K
Y2K compliant
Connectors and 10BaseT ports: RJ-45 connectors; two-pair category 3, 4, or 5
Cabling
unshielded twisted-pair (UTP) cabling
100BaseTX ports: RJ-45 connectors; two-pair Category 5 UTP
cabling
1000BaseX GBIC ports: SC fiber connectors, single mode or
multimode fiber
GigaStack GBIC ports: copper-based Cisco GigaStack cabling
Management console port: RJ-45 connector, RS-232 serial cabling
Indicators
Per-port status LEDs - link integrity, disabled, activity, speed, and
full-duplex indications
System status LEDs - system, RPS, and bandwidth utilization
indications
Warranty
Note All units include a lifetime return-to-factory warranty
57
Annexe 1E - Cisco Aironet 1200 Series
Item
Part Number
Specification
• AIR-AP1231G-x-K9
• Regulatory Domains: (x = Regulatory Domain)
• A = FCC
• E = ETSI
• I = Israel
• J = TELEC (Japan)
Customers are responsible for verifying approval for use in their individual countries. To
verify approval and to identify the regulatory domain that corresponds to a particular
country please visit:
http://www.cisco.com/go/aironet/compliance
Not all regulatory domains have been approved. As they are approved, the part numbers
will be available on the Global Price List.
Software
Cisco IOS Software
Data Rates
Supported
802.11g: 1, 2, 5.5, 6, 9, 11, 12, 18, 24, 36, 48, and 54 Mbps
Network
Standard
IEEE 802.11b and IEEE 802.11g
Uplink
Autosensing 802.3 10/100BASE-T Ethernet
Radio Module
Form Factor
Frequency
Band and
Operating
Channels
• 802.11a: CardBus (32-bit)
Americas (FCC)
2.412 to 2.462 GHz; 11
channels
ETSI
2.412 to 2.472 GHz; 13
channels
Israel
2.432 to 2.472 GHz; 9
channels
• 802.11b or 802.11g: Mini-PCI (32-bit)
Japan (TELEC)
2.412 to 2.472 GHz; 13 channels Orthogonal Frequency
Division Multiplexing (OFDM)
2.412 to 2.484 GHz; 14 channels Complementary Code
Keying (CCK)
Nonoverlapping
802.11g: 3
Channels
58
Wireless
Modulation
802.11g: Direct sequence spread spectrum (DSSS); OFDM
Receive
Sensitivity
(Typical)
802.11g:
6 Mbps: -90 dBm
9 Mbps: -84 dBm
12 Mbps: -82 dBm
18 Mbps: -80 dBm
24 Mbps: -77 dBm
36 Mbps: -73 dBm
48 Mbps: -72 dBm
54 Mbps: -72 dBm
Available
Transmit
Power Settings
(Maximum
power setting will
vary according to
individual
country
regulations)
802.11g:
CCK:
100 mW (20 dBm)
50 mW (17 dBm)
30 mW (15 dBm)
20 mW (13 dBm)
10 mW (10 dBm)
5 mW (7 dBm)
1 mW (0 dBm)
Range
802.11g: Outdoor
110 ft (34m) @ 54 Mbps
200 ft (61 m)@ 48 Mbps
225 ft (69 m) @ 36 Mbps
325 ft (99 m) @ 24 Mbps
400 ft (122 m) @ 18 Mbps
475 ft (145 m) @ 12 Mbps
490 ft (149 m) @ 11 Mbps
550 ft (168 m) @ 9 Mbps
650 ft (198 m) @ 6 Mbps
660 ft (201 m) @ 5.5 Mbps
690 ft (210 m) @ 2 Mbps
700 ft (213 m) @ 1 Mbps
802.11g: Indoor
90 ft (27 m) @ 54 Mbps
95 ft (29 m) @ 48 Mbps
100 ft (30 m) @ 36 Mbps
140 ft (43 m) @ 24 Mbps
180 ft (55 m) @ 18 Mbps
210 ft (64 m) @ 12 Mbps
220 ft (67 m) @ 11 Mbps
250 ft (76 m) @ 9 Mbps
300 ft (91 m) @ 6 Mbps
310 ft (94 m) @ 5.5 Mbps
350 ft (107 m) @ 2 Mbps
410 ft (125 m) @ 1 Mbps
Ranges and actual throughput vary based upon numerous environmental factors
so individual performance may differ.
Compliance
Standards
Safety
• UL 60950
• CAN/CSA C22.2 No. 60950
• IEC 60950
• UL 2043
Radio Approvals
• FCC Part 15.247
• RSS-210 (Canada)
• EN 300.328
• ARIB-STD 33 (Japan)
• ARIB-STD 66 (Japan)
• AS/NZS 4771 (Australia and New Zealand)
EMI and Susceptibility (Class B)
• FCC Part 15.107 and 15.109
• ICES-003 (Canada)
• VCCI (Japan)
• EN 301.489-1 and -17 (Europe)
• AS/NZS 3548
Security
• 802.11i, WPA2, WPA
• 802.1X
59
• AES, TKIP
Other
• IEEE 802.11g
• FCC Bulletin OET-65C
• RSS-102
Antenna
2.4 GHz Radio:
• Two RP-TNC connectors; 802.11g approved with:
–AIR-ANT1728, AIR-ANT1729, AIR-ANT2012, AIR-ANT2506, AIR-ANT3213, AIRANT3549, AIR-ANT4941, AIR-ANT5959, and AIR-ANT2410Y-R
Network
Management
BootP, Secure Shell (SSH) Protocol, Secure HTTP (HTTPS), Trivial File Transfer
Protocol (TFTP), FTP, Telnet, console port, Simple Network Management Protocol
(SNMP) MIB I and MIB II, CiscoWorks Resource Manager Essentials (RME), CiscoWorks
Software Image Manager (SWIM), CiscoWorks Campus Manager, CiscoWorks
CiscoView, and CiscoWorks WLSE
LEDs
Three indicators on the top panel report Ethernet activity and status, device operating
status, and radio activity and status.
Housing
Die-cast aluminum
Dimensions (H
x W x D)
1.660 x 6.562 x 7.232 in. (4.22 x 16.67 x 18.37 cm); add 0.517 in. (1.31 cm) height for
mounting bracket
Weight
1.725 lb (0.783 kg); add 0.4 lb (0.181 kg) for mounting bracket
Environmental
-4 to 122°F (-20 to 50°C), 10 to 90 percent humidity (noncondensing)
Memory and
Processor
IBM PowerPC405 (200 MHz)
16 MB RAM; 8 MB Flash memory
Input Power
Requirements
90 to 240 VAC ±10 percent (power supply)
48 VDC ±10 percent
Power Draw
6W maximum
Warranty
One year
Wi-Fi
Certification
60
ANNEXE 2: Extended Antennas Options
¾ Omnidirectional Antenna
An omnidirectional antenna works equally well in picking up signals from every
direction. Omni's make excellent general purpose and mobile antenna's. The longer an
omnidirectional antenna is, the better performance it will have.
¾ Parabolic or Dish Antenna
A parabolic antenna will give you the greatest range for your signal. The trade-off is that
they are more difficult to aim. A parabolic antenna is the obvious choice for a point-topoint fixed wireless installation.
¾ Yagi Antenna
A Yagi antenna is a good choice for point-to-point transmission, or point-to-multipoint
where the destination antenna are all close together or in a straight line.
¾ Patch Antenna and Sector Antenna
The most common use of the patch and sector antenna designs is as the customer antenna
in a point-to-point wireless broadband system.
61
Maximum Power Level (mW)
Regulatory Domain
Americas (-A)
(4 W EIRP maximum)
EMEA (-E)
(100 mW EIRP
maximum)
Japan (-J)
(10 mW/MHz EIRP
maximum)
Antenna Gain (dBi)
CCK
OFDM
5.2 (Omni)
100
30
9 (Patch)
100
30
10 (Yagi)
100
30
11 (Omni)
—
—
12 (Omni)
100
30
13 (Integrated patch)
100
30
13.5 (Yagi)
100
30
14 (Sector)
50
20
21 (Dish)
20
10
5.2 (Omni)
20
10
9 (Patch)
10
5
10 (Yagi)
10
5
11 (Omni)
—
—
12 (Omni)
5
1
13 (Integrated patch)
5
1
13.5 (Yagi)
5
1
14 (Sector)
1
1
21 (Dish)
11
—
5.2 (Omni)
10
10
9 (Patch)
10
10
10 (Yagi)
10
10
11 (Omni)
10
10
12 (Omni)
10
10
13 (Integrated patch)
10
10
13.5 (Yagi)
10
10
14 (Sector)
10
10
21 (Dish)
10
10
Maximum Power Levels Per Antenna Gain for IEEE 802.11g
62
ANNEXE 3:Flow On Demand for Video Throughput Control
Guillaume Milcent and Yang Cai, Carnegie Mellon University,, [email protected], published on AmI-Life
Workshop Proceedings, Spain, 2005 and to appear on LNAI 3864, 2006
Abstract. The bandwidth of the wireless network has been a bottleneck for live security video streaming. In this
paper, we introduce the concept of the bandwidth of human visual processing. We intend to match the human
information processing bandwidth with physical network bandwidth. An eye gaze-based interface is used to
optimize the video flow by adjusting the video resolution and compression to reduce the bandwidth by up to
75%. With the addition of a face-based compression method the bandwidth is reduced by up to 88%.
1
Introduction
The digital video cameras have been increasingly used in surveillance systems. For wireless network the
bandwidth is the limiting factor and displaying more than 10 video with high resolution, high frame rate
and high quality is currently a big issue [10][11]. This problem is even more significant for a large, wired
video surveillance system, which requires displaying hundreds of cameras. At the same time, Flow On
Demand (FOD) for video throughput control is one way to tackle this problem. It calculates the flow
required by the user according to the ways that the different screens are viewed. When a screen is viewed,
the display becomes larger and the compression is reduced. This way more details of the picture currently
displayed can be shown. Reducing the bandwidth has become a hot topic for research in recent years
because of the over increasing size of digital networks especially in the medical and defense sectors
[10][11][14]. Sending high resolution information for entire images is inappropriate for video systems.
Because of the human processing limitation, we will use high resolution information only at the current
point of interest. By matching the information content of the image to the information processing
capabilities of the human visual system, significant reductions in bandwidth can be realized, provided that
the point-of-gaze of the eye is known.
Eye gazed compression is a good way to decrease the bandwidth as described in [1] “Implementation of a
foveated image coding system for image bandwidth reduction” (Philip Kortum and Wilson Geisler). The
reference [1] describes a method using superpixels to reduce the bandwidth of a 256x256 grayscale picture.
This algorithm is based on eye gazed compression and the compression changes dynamically. Our
approach is different. The reference [1] proposes to have different levels of compression in a same image,
which is really efficient but linked to the perception of only one person. We propose a system based on
dynamic compression for a display that includes several screens. If several persons watch a display, full
global perception, full color images and large images are requested for surveillance purpose. This need of
accuracy forbids the use of usual compression methods based on picture interpolation [7][27][28][29]. We
have developed a system which matches the human vision area and provide displays equal to the size of
this area. The display’s resolution changes according to the eye movement. Finally, a face-based
compression allows us to significantly reduce the network traffic by compressing the background and
keeping the faces in high resolution.
2 Network Bandwidth Computation
Similar to a digital still picture camera, a network camera captures individual images and compresses them
into a JPEG format. The network camera can capture and compress, for example, 30 such individual
images per second (30 fps), and then make them available as a continuous flow of images over a network to
a viewing station. At a frame rate of about 16 fps and above, the viewer will perceive full motion video
(MJPEG). As each individual image is a complete JPEG compressed image, they will all have the same
guaranteed quality, determined by the compression and resolution level as defined for the network camera
or network video server. For a traditional video camera system, the user chooses in an administration
63
window the frame rate, the compression and the resolution size; these parameters remaining fixed during
the use of the system. These parameters are chosen according to the performance of the network. Our
objective is to adapt dynamically the parameters to optimize the visualization and the bandwidth use.
Traditionally, the screen of interest is displayed in a large window and with several small windows
bordering the largest one. Each screen could be displayed following given rules such as event switching or
time switching for these compression and resolution parameters would remain the same. In practice, a
video camera displayed on the small screen with a high resolution does not provide a significant
improvement for the viewer and is a waste of bandwidth. The correlation between image size, compression
and resolution level is illustrated schematically in Fig. 1. The size of the picture from the motion JPEG
flow increases with the resolution and the quality (less compression).
Image Size function of compression and resolution
40
Image Size
(Ko)
30-40
20-30
30
20
10-20
0-10
10
640*480
0
10
320*240
30
Image
resolution
50
Compression Rate (% )
70
90
160*120
Fig. 1. Image size Diagram AXIS-210
We create a model [27] of this function in order to optimize the link between human perception and image
size. We obtained three equations (1)(2)(3) to compute the bandwidth. The exponential equation (2) is
linked to the nature of JPEG compression. JPEG is designed for compressing full-color or gray-scale
images of natural, real-world scenes. JPEG removes redundant information from images which is why
equation (2) is based on the exponential: the non-redundant information could not be totally removed and
remains as the lowest size of the picture. The image resolution I in pixels and the compression rate
percentage CR are the main parameters used to compute the image size S in pixels and the resulting
bandwidth B (Mbits per second). First, the non-redundant information RI can be approximated by
equation (1).
RI = 3.44*10− 4 * I + 7.41
(1)
S = RI * (CR) −0.40
(2)
B = S * CN * FPS * SC
(3)
The constant SC represents the complexity of a scene. This constant is in most instances equal to 8. It is
used along with the number of cameras CN and the frames per seconds fps for each camera to carefully
estimate the bandwidth. Typically, for a set up of four cameras capturing 640*480 video at 30 fps with a
compression rate of 10%, the bandwidth required is 36.4 Mbit/sec. This is the reason why a system with
more than 12 cameras overtakes the capacity of traditional Ethernet cable (100 Mbits/sec). Equation (3)
will be used later to compute the bandwidth for a given image size.
64
3 Human Bandwidth Modeling
In the next section, we will answer three questions: What is the area accurately covered by our view? How
much visual information can a human process? How does the human vision behave outside of this small
area? The goal is to provide a video surveillance system which matches perfectly the capacity of the human
vision (also called the human bandwidth).
3.1 Details perception
Human vision is divided in two different processes [29]: a conscious and a subconscious process (Fig. 2). If
visual processing did not occur on a subconscious level then the act of seeing would become a labored,
arduous and inefficient burden. The peripheral visual processing system is particularly subconscious. This
division between the central, conscious role of vision and the unconscious peripheral role is especially
important for the field of orientation and mobility.
Fig.2. Two different levels of vision: conscious and subconscious [29]
The conscious vision includes the fovea zone and the attentive vision zone (Fig. 2). The high vision of the
fovea zone corresponds to an angle of vision of approximately 3°. Vision is considered to be poor except in
this zone. The observer directs the glance by a perpetual movement of the eye to direct the fovea axis
towards the part of the image retained for a fine analysis [2]. The close zone constitutes a zone of
monitoring whose interpretation allows the fast orientation of the eye towards detail chosen instinctively in
spite of a low acuity and without movement of the head. This research relates to certain details of the
image which require an intellectual act of interpretation.
Table 1. The vision angles, parameters for the Gaussian human distribution [30]
Category
Fovea zone
Attentive vision
Perceptive vision
Lateral vision
Horizontal
angle
+3°
+15°
+50°
+100°
Vertical up
angle
+3°
+8°
+35°
+50°
Vertical
down angle
-3°
-12°
-50°
-75°
To conclude this biological explanation of the human vision field, a normal person can not see any details
outside of the fovea zone. Which part of the screen can be seen in full detail? How much information is
included in this zone? Table 1 provides some measures of the vision area.
3.2 Limit to Human Vision & its Effect on Optimum Digital Image Resolution:
In general, the higher the resolution is, the better the image rendering is. The image would appear cleaner
and less jagged. The closer the dots in an image are, the more likely we can only see continuous patterns
and not see any dots. The eyes just blur everything together. The human eye only has a certain number of
light detectors in it. It is these sensors that convert the light into nerve impulses. Since these sensors are in
limited number, it makes sense that they can only handle a certain amount of information. When they have
taken in all they can, the brain goes to work and interprets the signals from the sensors, and also determines
what is likely to be between the sensors as well. The size of the peak of the cone does not change much,
and the eye is constantly working to focus the light on this area. In this area called the “fovea”, the eyes
have a densely packed bunch of sensors, which make up the central field of vision [2]. The further away the
screen is, the smaller the area it will cover at the end of the cone where it is located. Therefore, when an
65
object is viewed up-close, the eye's central field of vision is filled with the object. A lot of information
about the object is sent to the brain, and the detail of the object can be recognized. We will consider a
viewer 30 centimeters away from the screen. According to table 1 and some basic geometry, the total area
seen is equal to a large video screen (84cm diagonal). In the following demonstration, we will use this as
our screen size S. This screen will have a resolution of 1920x1080 pixels.
Fig. 3. Gaussian distribution of the human vision, (left) based on a uniformed distribution of the vertical up
angle [30], (right) based on a uniformed distribution of the vertical down angle [30]
The human eye distribution can be surrounded by two Gaussian distribution densities G (m1 , σ 1 )
and G (m2 , σ 2 ) because of the difference between the up and down vision The human distribution is not
uniforme, the vertical down angle value is 5 to 30% bigger than the vertical up angle value (Table 1). For
example, the attentive vision has an horizontal angle of 15°, vertical up of +8° and vertical down of -12°.
The parameters σ 1 and σ 2 are estimated from the cone of perception described in the Table 1. This
distribution can be considered as a filter for browsing the screen on the area of global vision A. The
value Z represents the percentage of the screen (described previously) seen at a distance of 30 cm. It is the
result from the density function fα in convolution with the area viewed H at the point M ( x, y ).
1 ⎛⎜ r 2 ⎞⎟
−
1
2⎜ σ 2 ⎟
f α ( x, y ) = ∫∫ 2 e ⎝ α ⎠ ds
A σ α 2π
Zα =
(4)
f α ( x, y ) * H ( x, y )
S
(5)
Z1 ≤ Z ≤ Z 2
(6)
The computation of Z1 = 21% and Z 2 = 25% show that human bandwidth has a small and limited
capacity. Only 25% of an 80cm diagonal screen is seen globally at a distance of 30cm. This is the reason
why the information sent to the viewer must be equal to his capacity and must be updated as the views
change. For a display that includes hundreds of cameras, the number of screens seen is usually over
estimated and network resources are wasted. Understanding the human behavior and capacity allows us to
create application which matches these characteristics and the network bandwidth.
3.3 Motion perception
Motion perception is an important aspect of the system, especially in video surveillance. It must be
enhanced to increase the efficiency of the global system. The peripheral retinal system is sometimes called
the "where" retina. It is involved with the subconscious control of human navigation. It is an old visual
system, having evolved long before central visual processing. The extreme far edges of the retina are purely
reflexive. When an object moves on the far retinal edge an immediate reflex swings the eyes in a direction
which aligns the moving object with the fovea. Closer in, the peripheral retinal tissue can "see" movement,
but there is no object recognition. When movement stops, the object becomes invisible.
66
According to this biological demonstration, the frame rate can not be reduced without a loss in the level of
the perception. We propose a system which will reduce the resolution for non-seen area and keep a high
frame rate for viewing comfort.
4 Bandwidth adaptation
4.1 Resolution adaptation
To demonstrate the performance of our system, we compare our CMU-system to existing systems. We have
established some real scenarios to analyze the bandwidth adaptation when switching between different
views. First, the user is watching a display with 4 equal screens. The screen selected is displayed in a full
screen mode for 10 seconds. Then the user is returned to the four screen view. This scenario (Fig. 4) is
tested with default static parameters: resolution (640x480) and compression (10%). For the dynamic
management, these two parameters will change according to the human behavior. As explained previously,
the number of frames per second is maintained at 30 to guarantee a full movement perception.
Fig. 4. 4 screens scenario, screen switching every 10 seconds
The first comparative test is based on static bandwidth management. The default resolution (640x480) and
compression (10%) are selected permanently. These parameters are chosen in order to obtain high
definition video when the user selects a screen in a full screen mode. The results (Fig. 5) were predictable.
Because none of the parameters evolve, the bandwidth does not change and 38Mbt/s is used permanently to
display 4 cameras in high definition.
Fig. 5. 38Mbt/s used for a static management (100%=100Mbt/s)
The second test is based on basic dynamic management of the bandwidth. When the user switches to the
full screen mode, the bit flow from the three other video cameras is disabled. It results a rectangle signal
(Fig. 6) with the high level corresponding to the four screen mode (38Mbt/s) and the low level to the full
screen mode (9Mbts). However, the bandwidth used globally is still too large to support over 12 video
cameras.
Fig. 6. 38Mbt/s – 9Mbt/s for a basic dynamic bandwidth management (100%=100Mbt/s)
67
In the third trial, we develop a full dynamic bandwidth management system in order to adjust the human
bandwidth to the network bandwidth. When the watcher switches to a 4 screens display, the image
resolution becomes smaller so that the image size diminishes according to the equation (3). As explained
previously, the viewer does not notice a change for over a certain resolution. This is why the low resolution
fits perfectly the small display. Globally the bit rate for the four screen display decreases from 38Mbt/s to
22Mbt/s (Fig. 7). The gain is approximately 43%.
Fig. 7. 22Mbt/s – 8Mbt/s for an evolved dynamic bandwidth management
4.2 Human Control
The bandwidth used for a one screen display can’t be reduced without decreasing the quality of the video.
The ratio between video resolution and video quality is optimal. In section 3, the human bandwidth is
presented as a limited resource, less than 25% of a screen being noticed that is the reason why it is useless
and expensive for a system to provide full detail quality for screens which are not watched. While using
this human property, we provide a system that increases the quality of the images viewed at the expense of
the quality of the other videos. An eye gazing device can locate the screen watched and transmit the
coordinates of the screen observed with an accuracy of a few centimeters. As explained in the chapter 3.3,
human attention is sensitive to motion. This is why the frame rate must remain maximal even for the screen
with low resolution and high compression.
Fig. 8. Real Time Eye Gazing
Fig. 9. Quality On Demand for a 4
screens video display
68
Fig. 10. Flow-chart Eye-gazing bandwidth control
One drawback that was observed in the first versions of the system was that the switching was too frequent
because of the eye’s reflex. For example, when screen 1 was watched just after screen 2 the modifications
that occurred on screen 2 is perceived as a movement and the eyes focus directly on the previous screen. To
avoid this phenomenon, a timer is incorporated to avoid this window wiper effect. To fully understand how
this drawback is decimated, the diagram (Fig. 10) describes the dynamic steps of the program. When using
the eye gazing, the bandwidth reduction is efficient. Globally, the reduction is over 50%. For our last test
system the bandwidth was about 22Mbt/s before the human control and 10Mbt/s after. In the Fig. 11, we
can see the difference between the evolved bandwidth management (22Mbt/s) and the human control
bandwidth management.
Fig. 11. 22Mbt/s – 10Mbt/s CMU dynamic bandwidth management (100%=100Mbt/s)
5
Face Based Image Compression
At this level, only screen, that are viewed are displayed in high resolution. Can we still reduce the size of
the picture while keeping the important information? In the field of video surveillance, it is interesting to
capture the face of the people moving in front of the video cameras and while the background is less
interesting but needed. We develop a video system including high definition faces and low resolution
backgrounds. First, a classifier is trained with a few hundred sample views of faces, which are called
positive examples. These are all scaled to the same size. Negative examples are arbitrary images of the
same size. After a classifier is trained, it can be applied to a region of interest (of the same size as used
69
during the training) in an input image. The classifier output is "1" if the region is likely to show the face
and "0" otherwise. To search for the object in the whole image one can move the search window across the
image and check every location using the classifier. The classifier is designed so that it can be easily
"resized" in order to be able to find the objects of interest at different sizes, which is more efficient than
resizing the image itself. So, to find an object of an unknown size in the image the scan procedure should
be done several times at different scales.
Fig. 12. Face based compression method overview
In comparison to the high resolution image (12.1KB) the composite image size is only 4.9KB. The image
reduction is about 60%, but the information is fully transmitted. The high definition face is added to the
low revolution background to preserve the global quality of the picture. When transmitting the coordinates
of the face found in the high definition picture, theses coordinates must be translated into the new
referential coordinates of the composite image. Equation (3) is used to compute the bandwidth for facebased compression video. To conclude, we will evaluate all the techniques described in this paper (Table
2).
Table 2. Bandwidth management comparative results
Management
Full screen
4 screen
Bandwidth
average
Reduction
Static
38Mbt/s
38Mbt/s
38Mbt/s
0%
Basic
Dynamic
9Mbt/s
38Mbt/s
24Mbt/s
39%
Evolved
9Mbt/s
22Mbt/s
16Mbt/s
57%
9Mbt/s
10Mbt/s
9,5Mbt/s
75%
4Mbt/s
5Mbt/s
4,5Mbt/s
88%
Human
control +
Evolved
Face based +
Human
control
+Evolved
70
6
Conclusion
The combination of the dynamic bandwidth management for the displays, the eye gazing control and the
face-based compression allow for a reduction in bandwidth of up to 88%. This value increases with the
number of cameras displayed. However, this system has limitations. The face-based compression is
efficient only if faces are detected on the video capture. This systems request a camera with several
resolution outputs to compute the composite image which is perceptually close from the original high
definition image.
7
Acknowledgement
We would like to thank Army Research Office (ARO) and Bombardier Transportation, USA for their
sponsorship. We are also in debt to Mr. Joshua Emerson and Mr. Lawrence R. Gallagher, director of Data
Communication for their professional support.
8
References
1
Implementation of a foveated image coding system for image bandwidth reduction, Philip Kortum and Wilson
Geisler, University of Texas Center for Vision and Image Sciences. Austin, Texas 78712
Visual Performance. In Bass, M. (Ed.) Geisler, W.S. and Banks, M.S. (1995), Handbook of Optics Volume 1:
Fundamentals, Techniques and Design, 2nd Edition, New York
Patent EP20030767118 Turner r Brough (us); Bruemmer Kevin j (us); Matatia Michael(us), methods and
apparatus for network signal aggregation and bandwidth reduction
Juday, R.D. and Fisher, T.E. (1989) Geometric Transformations for video compression and human teleoperator
display. SPIE Proceedings: Optical Pattern, Recognition, Vol. 1053, 116-123.
Warner, H.D., Serfoss, G.L. and Hubbard, D.C. (1993) Effects of Area-of-Interest Display Characteristics on
Visual Search Performance and Head Movements in Simulated Low-Level Flight. AL-TR-1993-0023. Armstrong
Laboratory, Human Resources Directorate, Aircrew Training Division, Williams AFB, AZ.
Wassel, H., Grünert, U., Röhrenbeck, J., and Boycott, B.B. (1990) Retinal ganglion cell density and cortical
magnification factor in the primate. Vision Research, 30, 1897-1911.
Weiman, C.F.R. (1990) Video Compression Via Log Polar Mapping. SPIE, Proceedings : Real Time Image
Processing II, Vol. 1295, 266-277.
Wilson, H.R., Levi. D., Maffei, L., Rovamo, J. and Devalois, R. (1990). The Perception of Form: Retina to Striate
Cortex. In L.S. & J.S. Werner (Eds.), Visual Perception: The Neurophysiological Foundations (pp 232-272). San
Diego: Academic Press.
Patent US20010865485 Peck Charles c (us); Mackay John d (us)
Eye gaze control of dynamic information presentation
R. Guerin, H. Ahmadi, and M. Naghshineh, "Equivalent capacity and its application to bandwidth allocation in
high-speed networks," IEEE J. Select. Areas Commun., vol. 9, pp. 968-981, 1991. Equivalent capacity and its
application to bandwidth allocation inhigh-speed networks
R. J. Gibbens , P. J. Hunt, Effective bandwidths for the multi-type UAS channel, Queueing Systems: Theory and
Applications, v.9 n.1-2, p.17-28, Oct. 1991
M. Sidi, W. Z. Liu, L Cidon, and I. Gopal, "Congestion control through input rate regulation," in Proc.
GLOBECOM '89, Dallas, TX, pp. 1764-1768.
Jon Kleinberg , Yuval Rabani , Éva Tardos, Allocating bandwidth for bursty connections, Proceedings of the
twenty-ninth annual ACM symposium on Theory of computing, p.664-673, May 04-06, 1997, El Paso, Texas,
United States
San-Qi Li , Song Chong , Chia-Lin Hwang, Link capacity allocation and network control by filtered input rate in
high-speed networks, IEEE/ACM Transactions on Networking (TON), v.3 n.1, p.10-25, Feb. 1995
Errin W. Fulp , Douglas S. Reeves, Bandwidth provisioning and pricing for networks with multiple classes of
service, Computer Networks: The International Journal of Computer and Telecommunications Networking, v.46
n.1, p.41-52, 16 September 2004
MeasuringBandwidth K Lai, M Baker - INFOCOM, 1999 - gunpowder.stanford.edu
Anwar I. Elwalid Debasis Mitra Effective bandwidth of general Markovian traffic sources and admission control
of high speed networks June 1993
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
71
18
19
20
21
22
23
24
25
26
27
28
29
30
real-time
foveated
multiresolution
system
for
low-bandwidth
video
communication
WS Geisler, JS Perry - Proc. SPIE, 1998 - svi.cps.utexas.edu
Patent US20040978903 Zimmerman Ofer (il); Stanwood Kenneth l (us); Bourlas Yair (us)
Method and apparatus for bandwidth request/grant protocols in a wireless communication system
Patent CA20032494956 Kandhadai Ananthapadmanabhan a (us); Manjunath Sharath (in); Bandwidth-adaptive
quantization
Patent WO2004IB51573 Sethuraman Ramanathan (nl), Method and apparatus for scalable signal processing
Patent WO2003EP10523 Riedel Michael (de); Neumann Roland (de), System and method for lossless reduction of
bandwidth of a data stream transmitted via a digital multimedia link
Patent US20000633217 Zahorjan John (us); Eager Derek l (ca); Vernon Mary k (us)
Bandwidth reduction of on-demand streaming data using flexible merger hierarchies
Patent CN19980807431 Wallerius John w (us); Walters Andrews j (us); Vastano John (us)
Method and apparatus for wireless communication employing control for confidence metric bandwidth reduction
Patent WO2003JP1 Kuroki Kenichiro (jp); Tajima Yuji (jp), Image processing display device and image
processing display method
Patent US20000568196 Bell Cynthia s (us), Microdisplay with eye gaze detection
Patent CA20032496389 Westphal Geoffrey a (us), System and method for image compression, storage, and
retrieva
Patent US20030684100 Walker Bradley k (us); Turnipseed John d (us); Method and apparatus for providing
interactive multimedia and high definition video
Institute for innovative blind navigation http://www.wayfinding.net/vsionsys.htm
http://www.rennes.supelec.fr
72
Was this manual useful for you? yes no
Thank you for your participation!

* Your assessment is very important for improving the work of artificial intelligence, which forms the content of this project

Download PDF

advertisement