基于光流法的视觉避障系统研究

Visual Obstacle Avoidance System Based on Optical Flow Method

本文针对无人机自主飞行过程中有效定位和识别障碍物的问题,研究了基于金字塔Lukas-Kanade光流法的视觉避障原理,在光流法基础上给出了障碍物检测方法及相应的避障策略。结合配有320 ×240像素摄像头的视觉模块和高性能嵌入式计算平台,构建了一套完整的实时视觉避障系统。实验表明,本文算法能够很好地识别障碍物并作出相应的避障动作,具有较好的实时性和鲁棒性。

To solve the problems of effective orientation and obstacle recognition in autonomous flight of Unmanned Aerial Vehicles, this essay has studied the visual obstacle avoidance principle based on Pyramid LK optical flow and the obstacle detection method as well as the relevant obstacle avoidance strategy with the foundation of optical flow. Combining with the 320 ×240 pixle camera- equipped visual module and high-performance embedded computing platform, we constitute a complete real-time visual obstacle avoidance system. Experiments show that the proposed algorithm can recognize obstacle well and react with relevant obstacle avoidance action, which has good real-time and robustness features.