A feature extraction algorithm based on improved Snake model for multi-pass seam tracking in robotic arc welding

Abstract

Visual sensor based seam tracking technology plays an important part in intelligentized robotic welding, and the feature extraction is an essential step. However, in multi-pass welding, the traditional algorithms are difficult to deal with the welding noises and irregular laser stripe patterns. Aiming at this problem, a feature extraction algorithm based on improved Snake model is proposed. By carefully redesigning the energy functional based on the unique gray-value distribution of the laser stripe, and optimizing the minimization procedure to accelerate the convergence, the Snake model is improved to a stripe extractor. The feature points are then extracted based on the curvature and the local moments of the stripe. Further, the algorithm is modified for butt welding seams, and a Kalman filter is introduced to handle the weaving process, so as to extend its application range. Experimental results suggest that the algorithm has good adaptability and robustness to multiple welding seams even under the interference of welding smoke, strong arc light, and spatter noises. Comparison tests prove the advantage of the algorithm in high stability and flexibility.