网格结构光视觉位姿测量系统设计

Design of Visual Pose Measuring System Based on Grid-Structured Light

针对位姿测量系统会增加被测物重量的问题,设计了网格结构光的视觉位姿测量系统。提出了一种基于二维平面靶标的网格结构光标定算法,利用靶标和结构光的共面性,同时标定网格结构光和相机,实现被测物表面三维点云的获取。为了验证网格结构光点云测量精度,进行了结构光标定实验以及点云精度验证实验,结果表明点云精度在0.04 mm以上。研究了点云配准的位姿测量算法。采用采样一致性初始配准(SAC-IA)算法对测量点云与参考点云进行粗配准,为后续点云精配准提供较好的初始位置。采用改进的迭代闭合点(ICP)算法进行点云精配准,引入随机采样一致(RANSAC)算法和法向量阈值过滤匹配点,以点到面的距离为误差函数,实现位姿测量。进行了点云配准算法对比实验和位姿测量实验,实验结果证明:被测物发生旋转运动时,该系统的最大误差不超过0.25°;平移运动时,最大误差不超过0.35 mm。

To address the problem that the pose measurement system will increase the weight of the measured object,a visual pose measuring system based on gridstructured light is developed.First,gridstructured light is used to obtain the 3D point cloud on the surface of an object.A gridstructuredlight calibration algorithm based on a 2D target is proposed.Based on the coplanarity of the target and the structured light,the gridstructured light and a camera can be calibrated simultaneously.To verify the accuracy of the point cloud obtained by the gridstructured light,structuredlight calibration and pointcloud accuracy measuring experiments are conducted.The results show that the point cloud accuracy is above 0.04 mm.Moreover,the pointcloud registration algorithm for pose measurement is studied.The sample consensus initial alignment(SACIA)algorithm is used for coarse registration because it provides a good initial position for fine pointcloud registration.The improved iterative closet point(ICP)algorithm is used for fine registration and random sample consensus(RANSAC)algorithm and a normal vector threshold are introduced to filter matching points.The distance from the point to the surface is used as the error function to perform pose measurements.Registration algorithm comparison and pose measurement experiments are conducted.The results prove that when the measured object is in rotational motion,the maximum error of the system does not exceed 0.25°and in translational motion,the maximum error does not exceed 0.35 mm.