Stereo visual image matching in remote welding

In order to achieve remote welding operation in complicated and dangerous environment and to improve the automation of remote welding robot, remote welding often use stereo visual technique to analysis task space and to do three-dimensional reconstruction for remote welding. Before arc striking, with the assistance of the stereo visual system and through offline programming system or remote teaching, generate robot remote welding task program at first, then manual control to do high-speed macro guide in free motor area and do low-speed positioning guide in limited motor area. And in order to achieve stereo vision task, image feature point matching become one of the most important parts. This paper designs an algorithm of the outer limit constraints fast matching, which based on pixel height difference, and greatly improved the accuracy and timeliness of matching.