Abstract: The styles and packaging forms of parts in the body shop of a vehicle manufacturing factory are complex and diverse, and the automatic picking of small metal parts that are stacked in an disorderly manner is a key link in achieving complete automation of parts loading. The article takes the automatic picking of a small metal part in an unordered stacked packaging as a case study and builds a verification system that includes robots, grippers, material boxes, and vision. The system focuses on three key steps involved in bin picking technology: part recognition, path planning, and transportation positioning. The research results show that the three key factors affecting the equipment running rate and bin cleaning rate are the selection of visual sensor type, the accuracy of part positioning after picking, and the design form and color of material bins. The identification of key factors affecting disorderly picking ensures the systematic reliability of disorderly picking of body parts, providing a reference for the application of disorderly bin picking in vehicle manufacturing workshops.

Key words: Machine vision; Disorderly bin picking; Body parts

摘要： 整车制造工厂车身车间零件样式和包装形式复杂多样，无序堆叠的小型金属零件自动 抓取是实现车身车间上件完全自动化的关键环节。文章以一个无序堆叠包装的小型金属零件 自动抓取为案例，搭建了包含机器人、抓手、料箱及视觉的验证系统，分别针对视觉抓取工 艺中涉及的零件识别、路径规划、搬运定位三个关键步骤进行验证。研究结果发现，影响开 动率和清箱率的三个关键因素是视觉传感器类型选择、抓取后零件定位精度、物料料箱设计 形式及颜色。影响无序抓取的关键因素识别，确保了车身零件无序抓取系统性可靠性，为整 车制造车身车间无序抓取应用起到借鉴作用。

关键词: 机器视觉；无序抓取；车身零件