Scene Simulation and Cooperative Target Detection During UAV Autonomous Landing

摘要

In order to overcome the problems of low guidance accuracy, poor autonomy and high experimental cost during the landing of UA V, a method of scene simulation and cooperative target detection in uav autonomous landing is proposed. In this method, QR code image with strong error correction ability is adopted as the cooperation target, Unreal Engine 4 (UE4) is used to build UA V landing simulation scene, and the UnrealCV plug-in is installed to enable UE4 to communicate with external programs. Finally, in the scene, the accuracy of yolov3 (You Only Look Once v3), a object detection algorithm based on deep learning adopted in this paper, to detect the cooperative target position on the runway is verified. The simulation results show that the algorithm can detect the landing target position with 100% accuracy and recall rate, ensuring the smooth completion of the whole landing process.