Small UAV Detection in Videos from a Single Moving Camera

The rapid application of Unmamned Aerial Vehicles (UAV) has triggered serious threats to public security, individual privacy, military security, etc. Thus, discovering unknown UAVs fast and reliably becomes more and more important. Among UAV detection techniques, the vision-based method is almost the lowest cost and the most easily-configured one. In this paper, we propose a UAV detection method based on a single moving camera to handle the problem for UAVs with fast moving speed. Firstly, we employ a motion estimation method to stabilize videos. Then, a low-rank based model is adopted to obtain the object proposals and finally, a CNN-SVM approach is used to further confirm real UAV objects. Two real UAV datasets are used to evaluate the proposed method and experimental results show that our method outperforms the baseline methods.