Abstract: In order to realize the automatic rust removal of large splicing welds on ship walls, a complex cross-welded joint type identification method based on artificial intelligence deep learning and three-line laser structured light was proposed.Based on the laser stripe feature difference of three-line laser to seven different types of cross welds, the features was extracted and used to learn by the ResNet deep residual learning model, and the model was trained using transfer learning, and an identification model with training accuracy rate close to 100% was obtained.The experimental verification shows that the model can effectively identify seven types of cross welds such as straight type, cross type, T type, left L, right L, left T, and right T, and can realize the cross weld type prediction during the operation of the wall-climbing rust removal robot in real-time.It lays a solid foundation for the path tracking, positioning and navigation of the rust removal wall-climbing robot.