Abstract: In view of industrial robots recognition and positioning for peg-in-hole assembly in flexible production unit,an uncalibrated visual positioning method which could adapt to the changeable product specification working conditions was proposed.It reduced the time consumption of complex calibration process and machine halt due to working condition change.To complete the extraction of target workpiece's minimum circumscribed rectangle angular point coordinates,image processing and character recognition algorithm based on depth image was designed.Through coordinate transformation between depth image and robot,the control parameter drive the robot was generated to complete coarse positioning.The target hole geometry parameters was extracted from the color image with Hough circles detection algorithm,and then the robot was driven to complete secondary positioning task through color image.The experimental result showed that the positioning accuracy of the proposed method was between 0.6～1.2mm,approximately met the requirements of subsequent assembly process,which was suitable for recognition and positioning of multi-specification and small batch peg-in-hole assembly process.

Key words: industrial robot, depth image, uncalibrated visual positioning, flexible production, assembly

摘要： 针对柔性生产单元工业机器人轴孔装配任务中工件的识别和定位,提出一种能适应小批量多规格装配任务的工业机器人无标定视觉定位方法,减少了因复杂的标定过程和工况变化造成的设备停机而消耗的时间。基于深度图像设计了图像处理和特征识别算法,完成了目标工件最小外接矩角点坐标的提取。通过深度图像数据和机器人之间的坐标转换关系计算得出运动控制参数,驱动机器人完成粗定位。运用霍夫圆检测算法从彩色图像中提取待装配圆孔的几何参数,进而驱动机器人完成二次定位。实验结果表明,该方法的定位精度稳定在0.6mm~1.2mm之间,基本满足后续装配过程对定位误差的要求,可用于多规格、小批量轴孔装配过程中目标的识别和定位。

关键词: 工业机器人, 深度图像, 无标定视觉定位, 柔性生产, 装配