Author(s)

Yalong Wang, Youwang Hu, Xiaoyan Sun and Feng He

Corresponding Author

Youwang Hu

ABSTRACT

In the application of welding seam recognition and tracking based on machine vision technology, many requirements are put forward for the recognition and detection of the welding seam trajectory in the image, which need to meet key indicators such as high precision, strong anti-noise interference, and high processing speed. The algorithm of extracting the edge and centerline of the weld is extremely important. Based on the characteristics of the weld image in the actual project, after preprocessing the target image such as grayscale, this paper proposes a method to determine the threshold value of the weld area based on the numerical characteristics of the range and standard deviation of the image pixel neighborhood. The foreground and background of the weld are accurately separated, and then the algorithm of extracting the edge of the weld using the pixel neighborhood range feature value, and then combining the least square method to extract the center line of the weld. In the MATLAB environment, a comprehensive performance comparison test between the algorithm and a variety of edge extraction algorithms shows that the algorithm has good performance in terms of mean square error (MSE), peak signalto-noise ratio (PSNR), and calculation speed. Experiments verify that the algorithm can extract the edges and center lines of welds stably, accurately and efficiently.

KEYWORDS

Weld seam tracking, image processing, neighbor range, edge extraction, weld seam