Research on Weld Recognition Based on MESR Adaptive Threshold Algorithm

To realize the automatic welding of welding robots using machine vision technology, it is necessary to accurately identify and track the weld trajectory. The key is to solve the position of the weld centerline through image processing. The quality of the weld image will be affected by noise, which greatly reduces the accuracy and efficiency of weld recognition. In order to eliminate the interference of the noise in the image on the extraction of the weld, this paper proposes the idea based on the Maximum Stable Extremum Region (MSER) algorithm, combined with the median filtering and morphological methods, and adaptively determines the optimal threshold of each image. The target area of the weld is segmented to extract the foreground from the background, and the center of the skeleton of the target area of the weld is fitted with a straight line as the position information of the weld. Through the processing experiments on a large number of weld seam image samples, the accuracy and effectiveness of the image segmentation algorithm and the straight line fitting algorithm are tested and evaluated. The results show that the algorithm in this paper has high weld seam recognition accuracy and anti-noise ability, and can provide image coordinate information of weld centerline for weld tracking and positioning.