Welding joints recognition based on Hausdorff distance

A recognition method for joints of welding work pieces based on Hausdorff distance is presented. According to the laser structured light vision method, a CCD camera was employed to capture the images of joints of welding work pieces with different kinds of stripes, and the central lines of laser stripes are extracted. Six standard joints templates are presented based on image shape features of welding joints. Hausdorff distance is adopted for measuring the match of laser stripes and standard templates in joints' shape library. When the distance between current stripe and the templates is the minimum of all, the corresponding joint type is detected. In order to accelerate the algorithm, the figures are standardized by Hough transform and Hotelling transform, and the dimension of templates searching space is reduced according to features of images. Experimental results could testify the validity of the recognition method.