This paper conducts a systematic study on the difficulty of accurately generating ortho-grayscale images of tunnels of any shape using point clouds obtained from mobile laser scanning. The study first obtains tunnel point cloud data using the self-developed TLSD （Tunnel Laser Scanning Detection） mobile tunnel detection system， and integrates distance threshold methods， image enhancement algorithms， clustering algorithms， and thinning algorithms to effectively handle redundant and abnormal information in the initial point cloud. On this basis， a method for generating tunnel ortho images combining longitudinal and circumferential calibration is proposed. The longitudinal calibration method uses the spacing between sleepers as a quantitative benchmark and combines intelligent labeling software to achieve effective longitudinal calibration of tunnel point cloud data. In the circumferential calibration algorithm， this paper first performs circumferential polygon fitting of the tunnel point cloud based on the Douglas-Peucker algorithm， followed by a projection center transformation. Finally， based on the equidistant division method proposed in this paper， the tunnel cross-section point cloud is simplified and calibrated to produce a grayscale image that matches the actual physical space， achieving efficient construction of ortho images of tunnels with any morphology. Test results show that the method developed in this paper can effectively improve the efficiency of measuring key geometric information and disease statistics in tunnels. The measurement accuracy of tunnel defects such as cracks and leaks reaches the centimeter level， providing accurate quantitative analysis data for tunnel service performance evaluation and management.