Research on spatial curve weld seam recognition and trajectory planning based on 3D+2D vision

Weld seam recognition is a crucial foundation for intelligent welding. To address the challenges of identifying complex spatial weld seams and planning welding paths, a method based on 3D visual image preprocessing and polynomial fitting for weld seam trajectory generation is proposed. An experimental platform for 3D vision was established to define the relationship between the weld seam coordinate system and the welding robots coordinate system, followed by camera calibration. Weld seam features were extracted through image grayscale conversion and binarization, while noise and interference points were filtered out using point cloud data to obtain the 3D information of the weld seam. A piecewise function fitting method was then applied to the extracted weld seam point cloud. The weld seam was divided into seven segments, each fitted using high-order polynomials. The resulting fitted path exhibited a maximum error of approximately 0.3 mm, with an average error below 0.2 mm, which falls within the acceptable tolerance range. This approach lays a solid foundation for the application of intelligent welding robots.