CAJ | 학술논문

针对在未知环境中移动机器人自主导航面临的障碍物实时检测问题，提出了一种基于Kinect深度技术的障碍物在线快速检测算法。在深度摄像机标定基础上，分析了摄像机运动造成视频流中的场景变化，重点研究了室内动态背景下的Kinect深度图像特征和障碍物在线快速检测。建立室内动态背景模型，采用背景减除法和连通体分析提取障碍物并归类，实现了对Kinect视频序列图像的在线快速检测。以轮式移动机器人为实验平台，验证了所提出算法的实时性、准确性和鲁棒性。
침대재미지배경중이동궤기인자주도항면림적장애물실시검측문제，제출료일충기우Kinect심도기술적장애물재선쾌속검측산법。재심도섭상궤표정기출상，분석료섭상궤운동조성시빈류중적장경변화，중점연구료실내동태배경하적Kinect심도도상특정화장애물재선쾌속검측。건립실내동태배경모형，채용배경감제법화련통체분석제취장애물병귀류，실현료대Kinect시빈서렬도상적재선쾌속검측。이륜식이동궤기인위실험평태，험증료소제출산법적실시성、준학성화로봉성。
In view of real-time obstacle detection problems facing the mobile robot autonomous navigation in the unknown and alien environment, this paper proposes a method of an on-line rapid obstacle detection algorithm based on Kinect depth technique. Based on the depth camera calibration, this paper analyzes scene changes in video streams when camera moves and focuses on studying characteristics of Kinect depth image and on-line rapid obstacle detections under dynamic background indoors. By building indoor dynamic background models, using the method of background subtraction and the analysis of connected sand body to extract and classify obstacles, on-line rapid detection of video image sequences come true. With the platform of wheeled mobile robot, the accuracy, robustness and real-time performance of proposed algorithm is tested.