Vehicles Detection of Traffic Flow Video Using Deep Learning

Abstract

The vehicle detection and tracking are important tasks in intelligent transportation system. The traditional methods of vehicle detection often cause the coarse-grained results due to suffering from the complex environments. YOLO is a pragmatic approach to multi-target detection with a simple and effective algorithm. This paper use YOLO to detect the moving vehicles and use a modified Kalman filter algorithm to dynamically track the detected vehicles, achieving overall competitive performance in day or night. The experimental results show the method is robust to occluding vehicles or congested roads and can obtain 92.11% average accuracy of vehicle counting at 2.55 fps speed.