Video Based Vehicle Detection and Tracking using Image Processing and Deep Learning

With the number of vehicles on the road expanding rapidly, it is imperative to maintain sophisticated vehicle systems and traffic management. To decrease the frequency of accidents, traffic management strategies that promote highway safety and pinpoint the causes of collisions should use vehicle identification and monitoring. A new elevated highway vehicle data set containing around 8,000 images extracted from videos with proper annotation is made from the perspective of Bangladesh which provides a complete data foundation for vehicle detection and tracking based on image processing and deep learning. Classification of the vehicle is done into 10 categories. For detection purposes, the YOLO v 5 model, MASK R-CNN model, and SSD(Single Shot Detection) have been used while YOLO v 5 performs better among all. For tracking, we have used the YOLO v 5 model, DeepSORT framework, and GOTURN method where YOLO v5 is ahead among all. Following identification and tracking, the Yolo V5 model is used to estimate speed and count the number of vehicles. To verify the suggested strategy, multiple highway-monitored recordings are taken in different areas of Bangladesh.