Aiming at the problems of low accuracy and information loss using single sensor for obstacle avoidance in UAVs (Unmanned Aerial Vehicles), a UAV autonomous obstacle avoidance method based on multi-sensor fusion was proposed in this paper. The improved Bayesian fusion algorithm is used to fuse the point cloud acquired by 2D lidar and depth camera to compensate for the areas that the 2D lidar cannot detect. At the same time, an octree map is generated based on the fused point cloud, and the UAV is replanned in real-time according to the updated map information to achieve autonomous obstacle avoidance in unknown environments. The experimental results show that the proposed method not only improves the accuracy of UAV perception of the surrounding environment, with the root mean square error of the fused point cloud is less than 0. 06 m, but also has good obstacle avoidance performance, with the distance between the UAV and obstacles is greater than 0. 5 m, ensuring its safe flight in unknown environments.