Abstract: To solve the problem of difficult identification and location of scattered rivets in the process of automatic nail feeding,a method of identifying and locating scattered rivets based on monocular vision was proposed.By studying the reflected light intensity of rivets,the gray-value distribution law of rivet rod was obtained,the rivet templates were constructed in a parametric way,and a rivet identification algorithm based on gray-value matching was proposed.The rivet's position was obtained by the combination of monocular vision and a laser displacement sensor.At the same time,to solve the problem of principle error in measuring rivet position under different depths by monocular vision,a measurement method of repeated positioning and photographing was studied.The maximum position deviation between the vacuum nozzle and the center of rivet rod was reduced from 7.86 mm to 0.87 mm.The experimental results showed that the rivet identification accuracy of gray-value matching algorithm was above 85% and it could adapt to changes in illumination from 30% to 70%.The longest nail grasping beat of the robot using the proposed method was 3.32 s,which could leave enough time to transport the rivet to the automatic drilling and riveting equipment and meet the requirement of the automatic drilling and riveting equipment.

Key words: monocular vision, rivet identification, template matching, position measurement, automatic nail feeding

摘要： 针对自动送钉过程中散堆铆钉识别与定位困难的问题,提出一种基于单目视觉的识别与定位方法。通过研究铆钉的反射光强获得钉杆的灰度分布规律,以参数化的方式构造了铆钉模板,并提出基于灰度匹配的铆钉识别算法。将单目视觉与激光位移传感器结合获取铆钉的空间位置,同时为解决单目视觉在测量不同深度下的铆钉位置存在原理性误差的问题,研究了一种重复定位拍照的测量方法,使真空吸嘴与钉杆中心的最大位置偏差由7.86 mm降低到0.87 mm。实验表明,灰度匹配算法的铆钉识别准确率在85%以上,并能适应30%~70%的照明亮度变化。使用该方法的机器人抓钉节拍最长为3.32 s,可留有充足的时间将铆钉输送至自动钻铆设备,满足自动钻铆设备的需求。

关键词: 单目视觉, 铆钉识别, 模板匹配, 位置测量, 自动送钉