One Stage Near-Ground Pothole Object Detection

To propose a single-stage object detection method for realizing real-time and accurate identification and detection of near-ground small pothole objects photographed by UAV. Through the special copy-paste technology of the pothole dataset obtained by the drone, combined with the key point detection branch, the Anchor and loss function are redesigned and improving the Soft-NMS. The automatic detection of near-ground small objects is finally realized. Results: On the pothole data set, the detection of the 4000 * 3000 size map can be completed within 6 s, and the AP on the test set can reach 85%. The experimental results show that the algorithm can not only accurately identify the potholes in the near-ground roads, but also meet the real-time requirements, and can be deployed on UAVs to detect near-ground objects such as potholes.