|M.Sc Student||Rubinstein Oren|
|Subject||Freeway Traffic Level of Service Detemining through|
Automated Video Feed Processing
|Department||Department of Civil and Environmental Engineering||Supervisor||Professor Barak Fishbain|
|Full Thesis text|
LOS - Level Of Service - is a standard index, set through the American Highway Capacity Manual (HCM) procedure for traffic facilities, in order to reflect the ability of a certain facility type to operate under various traffic conditions. LOS is graded using A-F scale: An index value of "A" would represent high performance, allowing high velocity and low delays of traffic, and index value of "F" would represent dis-functionality of the facility, resulting in a collapse of the traffic stream. Determining LOS is important for the macro understanding of road traffic condition in order to implement different operational traffic schemes and to relieve traffic jams and supply better traffic conditions for road users.
Traditionally, traffic facilities' LOS is determined based on measurements acquired by inductive loop detectors. These sensors, however, are pavement intrusive, and therefore force traffic disruption for installation and maintenance.
This study evaluates the use of automated data processing of video feed in order to determine a specific highway's LOS vs. the common technology used for the procedure. To do this, a software package that extracts traffic flow parameters from a video feed has been developed presenting theoretical and practical novelties, allowing real-time processing of the data.
To evaluate the package, 24 hours' worth of data from a set of five loop detectors, and concurrent video, by Halacha interchange of road 20 in Israel, was used. Both loop sensors and video cameras are part of the existing traffic control infrastructure used by Ayalon highways traffic control center.
The results show that the video processing scheme that was in use is capable of extracting traffic flow parameters from video feed. When evaluating LOS assessment through video, the output seems to be more robust than when LOS is estimated through loop detectors data, yet certain disadvantages has been spotted as well. The software package is available for research, allowing future research in the field with great flexibility.