【摘要】 相较于三维坐标测量仪、激光跟踪仪等传统测量设备，双目视觉检测是一种无接触的检测方式，不会对目标的运动产生干扰和破坏，且更为经济、灵活和方便，因此采用双目视觉技术对3-RRR+(S-P)并联仿生髋关节的位姿进行检测。为提高位姿检测的精度、鲁棒性和通用性，设计了一种方便提取且角点可自动辨识的自编码视觉标识物、提出了一种基于边缘交点坐标反馈式的角点检测算法，该算法主要包括HSV色彩空间下的标识物提取、Canny边缘检测、Hough直线检测、亚像素角点检测和三维重建等流程，将标识物角点检测简化为检测标识物边缘线段并求得交点，再经由亚像素角点检测算法进行提纯的过程。最后，通过双目相机与自编码视觉标识物的布置建立并联仿生髋关节的位姿检测系统，进行位姿检测实验并采用激光跟踪仪进行对比实验，通过分析与实验验证：该系统具备较高的精度、鲁棒性和通用性。更多还原

【Abstract】 Compared with traditional measuring equipment such as coordinate measuring machines and laser tracking instruments, binocular vision is a non-contact method, which will not disturb the movement of the target, and is more economical, flexible and convenient. In this paper, binocular vision technology is used to measure the pose of 3-RRR +(S-P) parallel bionic hip joint.In order to improve the accuracy, robustness and universality of pose detection, a self-coded visual marker is designed which can be easily extracted and automatically identified, and a corner detection algorithm is proposed based on coordinate feedback of edge intersections. The algorithm mainly includes marking extraction in the HSV color space, Canny edge detection, Hough line detection, sub-pixel corner detection and three-dimensional reconstruction. As a result, the process of corner detection of the marker is simplified to detecting the edge lines to obtain the intersection points, which are then purified through the sub-pixel corner detection algorithm. Finally, the pose measurement system was built and pose measuring experiment was carried out.A laser tracker was used for comparison.Through the analysis and experimental verification of this paper, the system proved to be accurate, robust and universal.更多还原