Traffic analysis studies are commonly performed to inform policy makers about the rate of traffic flow and the prevailing traffic patterns at key intersections. In urban environments, like Philadelphia, one common approach to conducting these studies is by temporarily deploying a video camera that acquires footage over a period of study. This video is then analyzed to produce data on the routes that cars take at the intersection over that time period.
Once the video data has been obtained it can be manually annotated to obtain the required traffic count or turn information. As one can imagine, this is a time and labor intensive process. Alternatively, there are a number of commercial entities such as Miovision that have automated or semi-automated systems that can perform this function. Planning agencies can contract with these services for a fee to perform these counts. What we are seeking to do in this project is to develop open source video analysis software that would reduce the cost of acquiring this data and, thus, make it possible to acquire more information about traffic flow in a timely manner. Ultimately we would like to develop algorithms that could be deployed on small, inexpensive embedded smart camera systems which could provide real time information about hundreds of intersections and give traffic planners and city operators an unprecedented level of information about traffic conditions.
|firstname.lastname@example.org||Taylor, C.J.||University of Pennsylvania||PI||Faculty - Tenured|
|Final Report||Vehicle_Detection_and_Counting_at_Road_Intersections.pdf||Aug. 28, 2018, 6:48 a.m.|
No match sources!