Cloud-Based Virtual Traffic Sensor Network with 511 CCTV Traffic Video Streams

University Transportation Centers Program2019PDFTech ReportJin, Peter J.Ge, YiZhang, Terry (Tianya)Wang, YizhouMartinez, JonathanRutgers University. Center for Advanced Infrastructure and TransportationUnited States. Department of Transportation. University Transportation Centers (UTC) ProgramUnited States. Department of Transportation. Office of the Assistant Secretary for Research and TechnologyUnited States. Department of Transportation. Federal Highway AdministrationRutgers University. Center for Advanced Infrastructure and TransportationNew JerseyUnited StatesAlgorithmsClosed circuit televisionCloud computingData analysisData fusionTraffic dataTraffic flowTraffic surveillanceVehicle occupancyVehicle trajectoriesVideoCAIT-UTC-NC56DTRT13-G-UTC28Final Report, 9/1/2018-9/30/2019The existing 511 New Jersey (NJ) traffic data feeds provide real-time travel time and event data. However, travel time is only one of the characteristics of traffic flow and the lack of flow and occupancy information make it difficult for transportation agencies to apply advanced TSM&O (Transportation System Management & Operations) strategies that need comprehensive inputs including travel time (speed), flow, and occupancy. This project developed a cloud-computing platform for state-wide traffic video data analytics. The platform will use the existing 511 closed circuit television (CCTV) traffic video streams to generate traffic flow and occupancy data to enrich the existing TRANSCOM data feed primarily containing travel time and event data. A longitudinal scanline based algorithm is developed to detect vehicle trajectories with minimal computational cost. Special treatment modules including static noise removal, clustering removal, and strand detection methods are developed to address the occlusions and lane-changing problems in the NJ511 CCTV traffic videos. The algorithm is deployed in an Amazon Web Service (AWS) based platform for large-scale processing of the large number of video feeds. The research outcome includes traffic flow data matched with TRANSCOM TransFusion data links. The evaluation results indicate promising accuracy, computational efficiency, and low cost of the proposed platform.137