In this paper, aiming at the timely collection of traffic information on roads, especially crossing roads, we propose a vehicle detection and tracking based on comprehensive features, which is realized by using UAV, so as to realize the collection of traffic information on crossing roads through the integration of "sky-ground". Real-time processing of video data received by the UAV, Otsu is used to continuously adjust the size of segmentation gray threshold of UAV in image processing at different heights. The frame difference method is used to realize the recognition and counting of vehicles in video, and then the real-time traffic flow on the road is calculated and extracted. The position, color and geometric features are integrated to find the position of the vehicle in the next frame, so as to realize the tracking of the target vehicle. Finally, the detection of vehicles on the current road and the extraction of their speed and track is realized, so as to realize the real-time monitoring of traffic state of the intersection. In the outdoor test, the method of vehicle tracking based on comprehensive features proposed in this paper was tested. The experimental results show that this method has higher accuracy than a single feature. Compared with machine learning and other methods to detect vehicles have a faster speed. Finally, the conclusion is that the method has high accuracy and speed, and its characteristics enable it to meet the needs of road vehicle detection under normal circumstances.

CICTP