Active Traffic Monitoring Through Large Scale Processing of Aerial Camera Array Networks

2020PDFTech ReportSarasua, Wayne A.Zhao, XiDavis, William J.Clemson UniversityCenter for Connected Multimodal Mobility, Clemson UniversityCenter for Connected Multimodal Mobility, Clemson UniversityUnited StatesAerial photographyAlgorithmsAutomatic brakingAutomatic vehicle detection and identification systemsData collectionInformation processingMappingTraffic dataTraffic surveillanceAerial Camera ArraysVehicle TrackingVehicle MappingComputer VisionDeep Learning69A3551747117Final Report (March 2018 \u2013 November 2019)This research focuses on automated processing of video from a persistent surveillance camera array to extract traffic data from a 25 square mile area. In previous research, we developed an automated traffic surveillance system capable of processing aerial camera array imagery to extract valid and useful traffic data for diverse applications. In this research, we continue to improve the system\u2019s capability by adding a novel multiple hypothesis tracking capabilities to improve vehicle tracking in congested traffic and adding a location identification algorithm to map vehicles throughout a network. Our evaluation has shown that the proposed system is capable of collecting speed, density, and volume data with an acceptable level of accuracy for many applications. The mapping of vehicles for a sample area was also successful. With further research, improved video preprocessing, enhanced resolution, and a higher frame rate, the accuracy of tracking vehicles can be improved significantly which will eventually allow the envisioned system to be able to accurately map the location of all vehicles throughout a camera array image sequence. A digital real-time \u201ctraffic map\u201d created by the envisioned system will provide a robust data set where data mining methods could be applied to enhance traffic management and provide data for a variety of traffic studies. A connected vehicle camera array application can open up plenty of possibilities in real-time traffic surveillance where erratic drivers can be identified automatically and warnings or even shut down commands can be sent to the erratic vehicles. The active sensing capability of such a system can potentially prevent some incidents from occurring thereby increasing safety and reducing incident induced traffic congestion.104