1100

Abstract: Obstacle detection is the key problem of UAV autonomous obstacle avoidance. This paper proposes a multi-obstacle detection method based on monocular vision. The proposed method uses YOLOv3 to detect the suspected obstacle regions. SURF algorithm is used to extract and match the feature points. The principle of visual expansion is used to confirm the obstacle. Multithreading programming technology is introduced to reduce the running time of the system. Simulation experiments are conducted to verify the effectiveness of the proposed method.