Seam Tracking And Depth Information Extraction Based On Robot Vision

By using an adaptive threshold SSDA fast image matching algorithm based on wavelet pyramid, the seam tracking of crawling robot experiment is taken under a single CCD. Also, since it is hard to obtain the depth information from only a single CCD, a binocular vision sensor is adopted. By the methods of a simplified binocular competition stereo depth measurement algorithm, the vertical distance between CCDS and work-piece has been calculated.The contents of the paper include these parts:Firstly, the hardware platform of weld seam tracking system of the crawling robot is established, and the welding seam recognition algorithm is studied. The weld image gathered from CCD is enhanced by the pretreatment of median filter and rectification. And according to the comparison of several fast image matching algorithms based on image pyramid, SSDA algorithm based on wavelet pyramid is adopted by the robot tracking system.Secondly, the seam-tracking experiment based on a single CCD is accomplished. Through the further optimization of the fast matching algorithm, an adaptive threshold SSDA fast matching algorithm based on wavelet pyramid is applied to acquire the seam deviation information. The results of tracking experiment based on PID controller show that the image processing algorithm has a good real-time and the robot has a good tracking effect.Finally, the experiment of binocular depth measurement is designed. Under an ideal experiment setting, the feather of images, gathered from left and right CCDs, is extracted by the way of threshold segmentation and canny edge detect operator. A simplified binocular competition stereo depth measurement algorithm is applied to find the match point and then get the depth information, which lays a foundation of robot precise motional control.