Research on UAV Obstacle Detection based on Data Fusion of Millimeter Wave Radar and Monocular Camera

Abstract—Obstacle avoidance detection of small UAVs has been a challenging problem because of size and weight constraints. In this paper, a fusion of MMW and monocular camera data is proposed for small UAVs obstacle detection systems. A MMW sensor is used to detect distance and angle and the image of obtacles capturing by the camera. Next, the target point information detected by MMW is calibrated into the image to complete the data fusion. Then, the optimized edge detection algorithm and image grayscale frequency saliency map are used to segment the obstacle area in images. The proposed method was evaluated by experiments in a real flying environment which consist of obstacles with textures and shadows. In the experiments, we successfully detect the shape of obstacles for complex situations. Obstacles with complex textures and shadows can be effectively detected, which shows that the method has good robustness.