针对清理空间非合作目标任务中如何获取目标的相对位置和姿态的难题,提出一种双目视觉位姿测量方法。首先,设计了基于双目视觉的位姿测量算法,利用弧支撑线段的方法快速检测目标表面的对接环,通过极线约束准则和光流法辅助跟踪算法建立复杂场景下的对接环检测和筛选机制;利用两次遍历法快速标记连通区域,加入面积和曲率约束,提取目标表面特征较突出的规则标志点。利用三维重建后的对接环平面和标志点建立目标坐标系,解算与世界坐标系之间的位姿关系。之后将算法移植到DSP6678信息处理平台,实现了对空间非合作目标的实时连续位姿测量。通过3组不同工况下的小卫星模型实验,验证了该方法的有效性。

Herein, a measurement method is proposed to obtain the binocular visual position and attitude of a target during the removal of a spatial non-cooperative target. First, an algorithm measuring the position and attitude based on binocular vision is designed. The algorithm detects the docking ring on the target surface based on arc-support line segments. Moreover, it establishes the detection and sifting mechanism of the docking ring in complex environments using polar line constraints and optical flow method. Further, regular marking points with prominent features on the target surface are extracted based on a two-pass algorithm coupled with area and curvature constraints. The measurement algorithm also builds the target coordinate system using the docking ring plane after the execution of a 3-D reconstruction method and regular marking points. Then, the pose relation under the world coordinate system is calculated. Hereafter, the algorithm is transplanted into the DSP6678 information processing platform, and the real-time continuous measurement of the positions and attitudes are realized for a spatial non-cooperative target. The effectiveness of the proposed method is verified using three small satellite model experiments under different working conditions.