Abstract: At present, in workshops, parts fixed on fixtures were commonly guided and gripped by pre-teaching the motion trajectories of robotic arms, resulting in poor human-machine interaction and low levels of intelligence. Machine vision, as an effective means of non-contact sensing and measurement, was applied in industrial scenarios. Based on monocular vision pose measurement technology, a multi-camera PnP positioning model and robotic arm guiding model were established to achieve spatial perception and guided grasping of the parts. Additionally, a pre-calibration method was proposed using multi-zero anchors to compensate system errors between robotic arm and vision module, which effectively decreasing the positioning and guiding errors of large parts.

Key words: machine vision, automatic positioning and guiding, perspective-n-point(PnP) camera model, multi-zero anchor

摘要： 目前车间中常通过预先示教机械臂运动轨迹的方式对夹具上位姿固定的零部件进行引导抓取，存在人机工程性差且智能化程度较低等问题。机器视觉作为非接触感知测量的有效手段已在工业场景中有所应用。基于单目视觉位姿测量技术建立多相机n点透视定位模型和机械臂引导模型，实现了对零部件的空间感知和引导抓取。提出一种用于补偿机械臂和视觉模块系统误差的多零位锚点的预标定方法，有效减小了大尺寸零部件的定位引导误差。

关键词: 机器视觉, 自动化定位引导, n点透视相机模型, 多零位锚点