Research and Design of Weld Seam Tracking Algorithm Based on Machine Vision and Machine Learning

This paper proposes an automatic weld seam tracking method based on laser vision and machine learning. The method comprises three main aspects: 1. Utilizing laser vision to obtain two-dimensional laser stripe images for quick acquisition of weld seam position information; 2. Implementing deep learning networks to perform image denoising processing, thereby eliminating the impact of noise (such as arc light and spatter) on weld seam feature extraction; 3. Using laser triangulation to detect weld points in real-time, thus improving the localization accuracy of the weld tracking process. The paper describes the image denoising method based on an improved Neighbor2Neighbor network, the laser streak target detection and tracking method based on the KCF tracking algorithm, and the weld point extraction method and real-time target tracking method. Corresponding experimental results and flowcharts are provided to verify the effectiveness and feasibility of the proposed method.