Research on Vehicle Detection Algorithm Based on Improved YOLOv5 and Kalman filter

The rise of Intelligent Transportation System (ITS) has put forward higher requirements for vehicle-related technologies, and vehicle detection and tracking is the basis for realising many functions of ITS. Therefore, this paper proposes a vehicle detection method based on the combination of improved YOLOv5 target detection algorithm and Kalman filter. On the basis of the original algorithm of YOLOv5, improvements are made for the unsatisfactory detection of small targets and the imbalance of the detected object categories, so as to reduce the leakage of detection, improve the accuracy rate, and realise fast and accurate vehicle detection, and combined with the Kalman filter for the stable tracking of the detected vehicles, which is then matched according to the Hungarian algorithm. The validation results show that the algorithm can better meet the real-time detection requirements of traffic scenes and has high accuracy, which has good practical value.