空间站机械臂位姿测量中合作靶标的快速识别

 空间站机械臂在完成辅助对接或者目标抓捕时,需要实时求取机械臂上的视觉传感器与目标上的合作靶标之间的位置和姿态,而其前提条件是合作靶标的快速识别。本文提出了一种合作靶标的快速识别算法。算法分为3大步骤:首先用Sobel算子和改进的非极大值抑制算法提取靶标图像的单像素边缘;然后将每条边缘分为两段,分别采用最小二乘法进行圆拟合,若两段拟合结果相似则该边缘属于圆形;最后根据圆形的大小在每个圆形周围开出一大一小两个正方形窗口,统计在两窗的补集内距离圆心较近的直线数量,若直线数量满足规定条件则认为是合作靶标。利用手眼相机、六自由度转台和合作靶标对算法进行了验证,实验结果表明该算法能在1.5 m的距离内准确识别合作靶标,且不受光照条件影响。合作靶标的识别算法快速、稳定、抗干扰能力强。

Fast recognition of cooperative target used for position and orientation measurement of space station's robot arm

When space station''s robot arm performs auxiliary docking or target arresting, position and orientation between the visual sensor fixed on robot arm and the cooperative target on object must be measured in real-time, and its prerequisite is fast recognition of cooperative target. A fast recognition algorithm of cooperative target is proposed. The algorithm consists of three steps. To begin with, Sobel operator and the improved non-maximum suppression algorithm are hired to extract single pixel edges in the picture of cooperative target. Moreover, every edge is split into two sections, and each section is fitted into a circle using least square circle fitting method. If two halves have similar fitting results, the edge belongs to a circle. Finally, we draw two square windows around each circle according to the circle''s radius, one big and one small. The number of straight lines that are in the complement area of the two windows and are close to the circle center is counted, and the cooperative target is identified if the number of straight lines suites the predetermined condition. Experiments using hand-eye camera, six-DOF turntable and the cooperative target are executed to test our algorithm. Results demonstrate that the proposed algorithm can accurately identify the cooperative target within a distance of 1.5 m regardless of lighting condition. In conclusion, the cooperative target recognition method is fast and stable and has strong anti-interference capability.