Object Detection of UAV for Anti-UAV Based on Improved YOLO v3

While unmanned aerial vehicle (UAV) technology brings convenience to modern life, it also leads to some problems. To achieve anti-UAV, the object detection technology of UAV is the key. YOLO v3, one of single-stage detectors, has the best detection performance for balancing the accuracy and speed through capturing deep and high-level features. In the basis of YOLO v3, this paper improves it to detect UAV more precisely and it's the first time to introduce YOLO v3 based algorithm to UAV object detection for anti-UAV. It adopts last four scales of feature maps instead of last three scales of feature maps to predict bounding boxes of objects, which can obtain more texture and contour information to detect small objects. At the same time, to reduce the calculation, the size of UAV in four scales feature maps is calculated according to input data, and then the number of anchor boxes is also adjusted. The experimental results demonstrate that the proposed approach achieves better detection accuracy and obtains more accurate bounding boxes of UAV with similar speed. Therefore, the proposed UAV detection technology can be applied in the field of anti-UAV.