Spatiotemporal Vehicle Tracking, Counting and Classification

Detecting and tracking vehicles from transportation surveillance videos is essential for applications ranging from traffic queue detection, volume calculation to incident and vehicle identification. However, challenges including object occlusion and shadowy condition often lead to poor detection and tracking performance. In this paper, we propose a new approach to tackle these challenges. It is done by performing background subtraction, vehicle segmentation, shadow detection and removal, and vehicle tracking. Further, we apply the results of the above approach to perform vehicle classification and vehicle counting.