面向工业零件的机器人单目立体匹配与抓取

为实现通用性强、快速、准确的工业机器人6自由度零件抓取，提出了一种基于单目视觉引导的零件3维抓取方法．首先，采用按倾角分层的Chamfer距离匹配算法建立图像与待匹配模板的相似度函数，并运用爬山法局部优化的遗传算法搜索最优匹配结果；然后，通过CAD（计算机辅助设计）模型建立离线3D模板库，将匹配算法拓展到适用于复杂结构零件的空间6自由度位姿检测；最后，由各坐标系间的矩阵转换和系统标定得到机器人的抓取信息，从而实现零件的3维抓取．实验结果表明，优化后的位姿检测算法在匹配速度和准确性上均有所提升，且基于该检测算法的机器人3维抓取实验的位置误差在2 mm以内、转角误差在2°以内，可用于工业智能机器人的零件抓取．

The Monocular Stereo Matching and Grasping of Robot for Industrial Parts

To realize universal, fast and accurate grasping of 6-DOF (degree of freedom) parts by an industrial robot, a 3D grasping method based on the guidance of monocular vision is proposed. Firstly, the similarity evaluation function between an image and a matching model is established by the Chamfer distance matching algorithm, in which the image is delimited according to direction angles. A genetic algorithm optimized locally by the hill climbing algorithm is applied to searching for the best matching result. Then, an offline 3D model library is established by CAD (computer aided design) model, and the matching algorithm is expanded to the spatial 6-DOF pose measurement of complex-structure parts. Finally, the grasping information is obtained by matrix transformations among all coordinates and the system calibration, so as to realize the 3D grasping of parts. The experiment results show that the optimized pose measurement algorithm improves the speed and accuracy of the matching process. With the proposed measurement algorithm, the position error within 2 mm and rotation error within 2° are achieved in the robotic 3D grasping experiments. So the measurement algorithm can be applied to the part grasping of industrial intelligent robots.