Proposes a fast autonomous obstacle avoidance algorithm applied in multi rotor UAV. Uses RGB-D camera to obtain RGB images and depth images, and applies hierarchical strategy for the description of depth images in order to distinguish obstacles and non obstacles; Searches for the obstacle avoidance path of UAV with regional searching on the field depth images, and then achieves the effect of autonomous obstacle avoidance. The algorithm is tested on ROS, and the results show that the proposed method is less time-consuming, high accuracy and reliability.