<p>The research work is conducted to solve the problem of image recognition and process planning related to weld seam tracking for multi-layer and multi-pass welding of thick-walled workpieces with large V-grooves. Firstly, based on multi-layer and multi-pass welding process experiments, a welding layer and weld bead planning strategy is proposed. Then, a BP neural network model optimized using the grey wolf algorithm is proposed. This model is used to predict the cross-sectional area of the weld seam and the depth of sidewall penetration based on welding speed and wire feeding speed. To solve the problem of strong arc light interference and large spatter in multi-layer and multi-pass welding processes, which make weld seam tracking image processing difficult, a lightweight laser stripe image recognition algorithm has been developed. This algorithm employs histogram equalization for image enhancement, combined with multi-frame and operation and connected component filtering. Additionally, a strategy utilizing the CenterNet-seUnet model to assist in weld seam recognition has been proposed. Then, a strategy for calculating the cross-sectional area of the subsequent weld seam is proposed based on the contour of the formed weld seam obtained from image recognition and the ideal contour of the subsequent weld seam. The calculation results are used for welding process planning and robot control for subsequent weld seam formation strategies. Finally, the effectiveness of the proposed algorithm in this paper is verified through weld seam tracking experiments. The experimental results show that the method proposed in this paper can effectively meet the requirements of MLMP welding control for robots and can simultaneously meet the requirements of welding quality and welding efficiency.</p>