Segmentation-based Geometry Feature Estimation Framework for Autonomous Welding Robots in Curved Ship Block Manufacturing

In ship manufacturing, welding robots are used for welding automation. However, welding robots are difficult to use in curved block environments due to structural complexities. To solve this problem, accurate recognition technology of curved ship block welding environment for welding robots is required. This study proposes an objective segmentation-based geometric feature estimation framework for autonomous welding robots in curved ship block manufacturing. The geometric feature estimation of curved blocks was performed using a vision sensor, and the curvatures were estimated through the Hessian matrix, and the slopes between the parts were estimated through RANSAC (Random Simple Consensus) and SVD (Singular Value Decomposition). Finally, the gaps between the parts were estimated through segmentation and geometric feature extraction combined with DWT (discrete wavelet transform) and soft-argmax. The geometric feature estimation performance was evaluated by comparing the actual values with the estimated values through a test bed. The results of the study demonstrate that welding robots can achieve accurate and efficient welding in curved ship block. Future work will apply the proposed geometric feature estimation framework to real curved ship blocks.