Abstract: To automatically grind and polish weld bead on large-scale structure parts, grinding/ polishing robot should not only be capable of moving and polishing, but also accurately tracking and measuring the weld bead in complex industrial environment. The collected data is useful for weld-bead grinding and polishing. The vision system for the grinding and polishing robot in this paper consists of a double charge-coupled device(CCD) camera and a laser, both fixing on the robot. According to the brightness distribution of structured light, a region of interest(ROI) positioning algorithm for image pair is proposed. By using the new algorithm, the calculating efficiency is largely improved. Gaussian differential method related to the image columns is proposed, and the typical lines of structured light can be extracted from noise-polluted images. Rough location of the weld bead edge is realized by finding maximum and minimum difference values, while the exact bead edge is accurately and rapidly extracted based on definition of distance threshold. Thus, the geometric information of the weld bead can be obtained. The results of image processing experiments and grinding/ polishing experiments show that the visual system is reliable and stable, and vision algorithm is rapid and accurate. The system would lay foundation for automatically grinding and polishing large-scale structure parts in a complex industrial environment.

Key words: Grinding/polishing, Structured light, Vision algorithm, Weld bead, collaborative manufacturing, constraint characteristics, decision making, multiobjective optimization, networked manufacturing