Design of Traffic Flow Analysis System Based on UAV Video

The rapid development of computer vision system has provided great convenience for traffic industry research. In order to apply computer vision to the video shot by unmanned aerial vehicle (UAV), this paper conducts an in-depth study on background separation, image smoothing, and morphological processing; introduces the principle and algorithm used in the processing; and analyzes the advantages and disadvantages of each algorithm. Finally, the most suitable method for UAV video is determined and applied to the intersection of Huizhan road and Fenghuangzhong road in Nanchang city, and the vehicles were successfully detected. Results show that the proposed method is suitable for detecting vehicles. The traffic flow is processed and analyzed by means of computer vision, which is more accurate and rapid than manual recognition, and can provide more correct, complete, and detailed information to investigators.