\r梅江平，王 浩，张 舵，闫 寒，李 策\r
\r
AuthorsHTML:\r梅江平，王 浩，张 舵，闫 寒，李 策\r
\r
AuthorsListE:\rMei Jiangping，Wang Hao，Zhang Duo，Yan Han，Li Ce\r
\r
AuthorsHTMLE:\rMei Jiangping，Wang Hao，Zhang Duo，Yan Han，Li Ce\r
\r
Unit:\r天津大学机构理论与装备设计教育部重点实验室，天津 300350\r
\r
Unit_EngLish:\rKey Laboratory of Mechanism Theory and Equipment Design，Ministry of Education，Tianjin University，Tianjin 300350，China\r
\r
Abstract_Chinese:\r\r在\r高速并联机器人视觉分拣系统中，由于安装误差等原因，视觉系统与传送带以及机器人与传送带之间会存在一定的角度偏差，导致目标物体在分拣系统中的定位精度较低，使机器人末端不能精确抓取动态目标\r．\r而且传统的单目视觉定位及跟踪算法存在精度低、运算速度慢等问题，影响视觉分拣系统的计算速度，降低系统的实时性\r．\r针对以上问题，提出了一种基于圆特征的单目视觉定位与传送带标定结合的动态目标跟踪算法，首先不需考虑相机坐标系与传送带坐标系的角度偏差，拍摄一个圆形目标，并以光心为顶点构造直椭圆锥的几何成像模型\r．\r在此基础上利用其几何关系求取圆心在相机坐标系下的深度信息，再结合传送带标定与张正友标定方法，确定传送带编码器的比例因子以及机器人坐标系和传送带坐标系之间的角度偏差，并实时跟踪圆形目标运动过程中圆心在机器人下的位置\r．\r最后，对所提算法和传统算法分别进行实验对比，结果表明，该动态跟踪算法通过圆锥投影的几何关系推导圆心的空间坐标，避免了传统复杂的非线性方程求解，定位精度和计算速度大幅提高\r．\r该算法消除了传统算法的角度偏差，平均位置误差为\r0.82mm\r，具有较高的定位精度\r．\r运行中系统耗时较短，运行效率得到了有效提高，且误抓率和漏抓率分别为\r0.2\r%\r和\r0\r，具有很高的实用价值\r．\r\r
\r
Abstract_English:\r\rDue to installation errors and other reasons\r，\ra certain angle deviation has existed between the visual system and the conveyor belt and between the robot and the conveyor belt\r，\rleading to a low-positioning accuracy of the target object in the visual system\r．\rThis has made the robot end unable to catch its dynamic target accurately. Moreover\r，\rthe traditional monocular vision positioning and tracking algorithm has had some problems\r，\rsuch as low accuracy and slow operation speed\r，\raffecting the calculation speed of the visual sorting system and reducing real-time system performance\r．\rAddressing these problems\r，\ra dynamic target tracking algorithm\r，\rbased on a circular feature combining monocular vision positioning and a conveyor belt calibration is proposed\r．\rA circular target was shot by the camera without consideration of the angle of deviation between the camera coordinate system and the conveyor coordinate system\r，\rand a geometric imaging model of the straight elliptical cone\r，\rwith its optical center as the apex\r，\rwas constructed\r．\rOn this basis\r，\rthe depth information of the center of a circle in the camera coordinate system was obtained using its geometric relationship\r．\rCombining conveyor belt calibration and the Zhang Zhengyou calibration method\r，\rthe ratio factor of the conveyor belt encoder and the angle deviation between the robot coordinate system and the conveyor belt coordinate system were determined\r．\rThe position of the center of the circle under the robot was then tracked in real time\r．\rFinally\r，\rexperiments were carried out on both the proposed algorithm and traditional algorithms\r．\rResults showed that this dynamic tracking algorithm can deduce the spatial coordinates of the center of a circle through the geometric relationship of conical projection\r，\rwhile avoiding solution of the traditional complex non-linear equation\r．\rThe positioning accuracy and calculation speed were improved greatly\r．\rThis algorithm can eliminate the angle deviation of traditional methods\r．\rThe average positioning error was 0.82mm\r，\rachieving a higher positioning accuracy\r．\rThe system consumed less time in the operation\r，\rthe operation efficiency was effectively improved\r，\rand the rate of wrong catch and missing catch were 0.2\r%\r and 0respectively\r，\rwhich displayed good practical value\r．\r\r
\r
Keyword_Chinese:单目视觉定位；高速并联机器人；相机标定；传送带标定；动态跟踪\r

Keywords_English:\rmonocular vision positioning；high-speed parallel robot；camera calibration；conveyor calibration；dynamic tracking\r
\r


PDF全文下载地址:http://xbzrb.tju.edu.cn/#/digest?ArticleID=6403