UAV Obstacle Detection Algorithm Based on Improved ORB Sparse Optical Flow

With the rapid development of drone technology, the research on autonomous autonomous obstacle avoidance technology has become a hot spot. Aiming at the problem of real-time and accuracy of obstacle detection in complex environment, this paper proposes an obstacle detection algorithm based on improved ORB sparse optical flow. This paper deeply studies the principles of traditional HS and LK optical flow algorithms. The improved ORB algorithm is used to detect feature points, which improves the real-time and robustness of feature point extraction. The LK optical stream is then used to establish a correspondence between feature points before and after the frame image. The image pyramid model is used to solve the problem of unstable optical flow detection in large displacement images. The paper algorithm was verified by semi-physical simulation experiment and real flight experiment of drone. The image processing time of two frames is less than 100ms, which can meet the real-time requirements of drone flight.