Robust features extraction for lap welding seam tracking system

A robust method is presented for features extraction based on structured light images of lap weld joint. On the basis of image processing such as smoothing, adaptive thresholding and thinning, two steps are taken to extract the features of the images. First, the position of the features is located roughly by dividing the centerline into segments and the intercept and slope of each segment obtained using least-squares approximation; afterwards, the range of feature points can be detected by comparing intercepts between segments. Then image features are extracted precisely via the intersections of three straight lines fitted by least square. Experimental results verified the proposed algorithms.