An efficient hybrid approach for vehicle detection and tracking

Real-time vehicle detection and tracking is a major research problem in traffic surveillance systems as it aims to improve road safety, optimize traffic flow, and increase transportation efficiency. Traffic surveillance systems typically collect and analyze live traffic footages from multiple cameras, to provide valuable and rich information compared to other sensor-based approaches. However, real-time analysis of traffic footage can be challenging due to the complexity of detecting and tracking objects in the fields. In this paper, we propose an efficient and robust framework for vehicle detection and tracking that is suitable for low-resource and edge devices without compromising precision and accuracy. Our approach combines the Background Subtraction method (BS), an efficient moving object detector, with a deep CNN-based detector in a hybrid detection approach.