Abstract: In the production of lithium battery packs, the accurate positioning and high positioning accuracy and speed of the pole weld points are crucial for the quality of the welding. This paper proposes an algorithm for the positioning of battery module pole weld points based on machine vision. Firstly, the acquired images were preprocessed by grayscale, bilateral filtering and denoising, gamma correction, adaptive binarization and morphological operations. Then, rectangle and circular contours were filtered and separated through an improved Mean-Shift clustering algorithm. After sub-pixel processing, the random sample consensus(RANSAC) algorithm was used to fit four straight lines, and an improved center-based clustering algorithm was used to fit circles. Finally, the weld point coordinates were calculated by using the obtained linear and circular equations. The experiments have demonstrated that the algorithm possesses high positioning accuracy and speed, which satisfies the requirements of industrial production.
Keywords: machine vision; weld joint positioning; clustering algorithms; straight line fitting; circle fitting
合肥工业大学学报自科版