Abstract: Objective Aiming at the application requirements of high efficiency and high quality in robotic automated welding, this paper proposes a method of weld groove feature extraction based on structured light sensor scanning and spatial plane equation intersection. Methods First, a structured light camera is used to capture images and obtain the 3D point cloud data of the workpiece to be welded. Second, the DBSCAN algorithm and voxel grid method are applied for point cloud filtering and downsampling. Then, the random sample consensus (RANSAC) plane segmentation with quadratic approximation is adopted to obtain the surface feature equations of the weldment and groove. Furthermore, the principal component analysis (PCA) combined with the 2D convex hull (Melkman) algorithm is used to extract the feature equations of the boundary planes at both ends of the weld. Finally, the 3D information of weld groove feature points and the geometric features of the groove are extracted by means of the intersection of multiple spatial plane equations. To verify the feasibility and adaptability of this method, experiments on groove feature extraction are carried out by using common V-groove butt joints, lap-grooves and T-grooves respectively. Results The experimental results show that this method can accurately extract the 3D information of feature points at the weld groove and the geometric shape of the groove, with a recognition error less than 1 mm. Conclusion The method of weld groove feature extraction proposed in this paper can meet the high-precision and automation requirements of robotic welding.