With the iterative progress of science and technology, the traditional manufacturing industry is gradually shifting to centering on industrial robots as the core to carry out intelligent manufacturing. Due to the characteristics of large thin-walled, diverse structures, and varying sizes of rocket storage tank parts, most of the current production methods are manual processing, which is relatively inefficient. In order to solve the above difficulties, taking the pre-welding treatment method of rocket fuel storage tank parts as the research object, a robot automatic grinding technology scheme from scanning, planning to execution is designed. The pose relationship between the point cloud in the machining area and the robot base coordinate system was obtained by means of hand – eye calibration and coordinate transformation. A point cloud splicing method was proposed to extract the edge surrounding box, and corner detection was carried out using the window algorithm based on the edge curvature. Finally, the robot automatic grinding experiment of pre-welding processing of parts was completed according to the machining trajectory planning results, ensuring the grinding quality and efficiency. It has obtained a good application effect.