Much more attentions have been attracted to the inspection and prevention of unmanned aerial vehicle (UAV) in the wake of increasing high frequency of security accident. Many factors like the interferences and the small fuselage of UAV pose challenges to the timely detection of the UAV. In our work, we present a system that is capable of detecting, recognizing, and tracking an UAV using single camera automatically. For our method, a single pan–tilt–zoom (PTZ) camera detects flying objects and gets their trajectories; then, the trajectory identified as a UAV guides the camera and PTZ to capture the detailed region image of the target. Therefore, the images can be classified into the UAV and interference classes (such as birds) by the convolution neural network classifier trained with our image dataset. For the target recognized as a UAV with the double verification, the radio jammer emits the interferential radio to disturb its control radio and GPS. This system could be applied in some complex environment where many birds and UAV appear simultaneously.

中文翻译：

---

基于轨迹和图像的无人机检测与识别

随着安全事故频率的增加，无人机（UAV）的检查和预防越来越受到关注。无人机受干扰、机身小等诸多因素对无人机的及时发现提出了挑战。在我们的工作中，我们提出了一个能够使用单个摄像头自动检测、识别和跟踪无人机的系统。对于我们的方法，单个平移-倾斜-变焦 (PTZ) 摄像机检测飞行物体并获取它们的轨迹；然后，识别为无人机的轨迹引导摄像机和云台捕捉目标的详细区域图像。因此，使用我们的图像数据集训练的卷积神经网络分类器可以将图像分为无人机和干扰类（例如鸟类）。对于双重验证识别为无人机的目标，无线电干扰机发射干扰无线电干扰其控制无线电和GPS。该系统可以应用于一些鸟类和无人机同时出现的复杂环境。