无人机航拍视频中的车辆检测方法

Vehicle Detection Method in UAV Aerial Video

利用无人机航拍获取交通信息具有灵活性好、机动性高等优势,能在突发事件、应急处理等环境发挥重要作用。提出一种针对无人机航拍视频进行车辆检测的方法。针对无人机视频的偏转问题,采用基于SURF算法的图像配准方法,将无人机视频的偏转根据标准帧进行矫正。通过基于像素的自适应分割算法将背景和运动车辆分隔,检测出视频中运动的车辆目标。实验表明:本方法匹配的准确率达到90%,能在较大的场景范围内检测出运动车辆,召回率(Re),精度(Pr),假阴性率(FNR)达到82.3%,90.1%,17.3%。

The use of UAV aerial photography to obtain traffic information has many advantages like a flexible high mobility, comprehensive information and so on, which plays a vital role in the emergency response, anti-terrorism/battlefield and other environments. This paper proposes a method for vehicle detection in UAV aerial video. Aiming at the problem of UAV video image deflection, the image registration method based on SURF algorithm is used to correct the deflection of UAV hatching video according to the standard frame. The background and the moving vehicle are separated by the pixel-based adaptive segmentation algorithm to detect the moving vehicle in the video. Experiments show that the matching accuracy of our method rate up to 90%, and can detect the movement of vehicles accurately in a large range of scenes, whose recall（Re）, precision（Pr）, false negative rate（FNR） are 82.3%, 90.1%, 17.3%.