CAJ | 학술논문

自主移动机器人的导航要求能实时规避障碍物。为解决该问题，提出一种基于光流的自主移动机器人避障系统。将机器人在运动过程中通过摄像头拍摄的连续图像序列作为输入，使用金字塔Lucas-Kanade方法在角点处提取光流信息求解延伸焦点和接近时间，并通过改进的平衡策略估计障碍物可能存在的位置。实验结果表明，该系统在静态场景中能正确地检测并规避宽度大于10 cm，且不完全遮挡摄像头视角的障碍物，可行性较高。
자주이동궤기인적도항요구능실시규피장애물。위해결해문제，제출일충기우광류적자주이동궤기인피장계통。장궤기인재운동과정중통과섭상두박섭적련속도상서렬작위수입，사용금자탑Lucas-Kanade방법재각점처제취광류신식구해연신초점화접근시간，병통과개진적평형책략고계장애물가능존재적위치。실험결과표명，해계통재정태장경중능정학지검측병규피관도대우10 cm，차불완전차당섭상두시각적장애물，가행성교고。
Since real-time obstacle avoidance is a key part of mobile robot autonomous navigation, an autonomous mobile robot obstacle avoidance system based on optical flow is proposed. This system’s input is the image sequence grabbed by the camera on the robot in motion. The optical flow information is extracted by Pyramid Lucas-Kanade method to calculate focus of expansion and time to contact. The possible position of the obstacle is estimated with improved balance strategy. Experimental results show that this system can detect and avoid obstacles with the width larger than 10 centimeters but not cover the camera view correctly, which validates the method’s feasibility.