The safety detection of robot in hydraulic tunnel is an effective means to solve the inspection of inclined shaft section of diversion tunnel. In this paper, the crawling robot is equipped with three-dimensional laser scanner and high-definition camera equipment for structural inspection, and the defect information of the diversion tunnel section is obtained through three-dimensional point cloud processing and defect data splicing. Combined with the defect data processing results, the structural defect characteristics and distribution law of the diversion tunnel are analyzed to realize the inspection and analysis of the irregular shape, uneven distribution and coexistence of multiple defects of the diversion tunnel. The results show that the defects in the diversion tunnel include cracks, leakage and reinforcement leakage, and cracks are the main defect type of the diversion tunnel, and the number of defects in the upper flat section is significantly more than that in the inclined shaft section. The cracks in the upper flat section are mainly distributed at the top of the tunnel, and most of the cracks are horizontal cracks perpendicular to the tunnel axis. Fractures in inclined shaft section are evenly distributed in all positions around the tunnel, and most of them are inclined fractures. It can be seen that there are great differences in the number of defects and fracture characteristics between the upper flat section and the inclined section. In the later safety inspection process, we should mainly pay attention to the expansion of fractures.