A geometry based feature detection method of V-groove weld seams for thick plate welding robots

The precise positions of feature points in the image space are crucial for tracking systems of welding robots. And hence the quality of feature detection directly affects the accuracy of the motion of the welding robots. Moreover, a low computational cost is also needed in a real-time tracking system. This paper presents a method which effectively detects the features of the V-groove weld seam based on their geometries. A camera and a cross-line structured light are constructed to be used as the visual sensing system. Firstly, the skeletonization has applied to the image. Subsequently, Hough transform is used to find the main line of the light stripe. Secondly, The farthest point algorithm is used to estimate the bottom feature point of the weld seam. Finally, the left and right feature points are estimated by the least squares line fitting method. The experimental results show that the features can be precisely recognised by the proposed method. In addition, the computational cost of the proposed method is not expensive, and thus it satisfies the requirement of a real-time system.