Underground pipe inner surface defect detection is constrained by space accessib ility.In this work we propose an in-pipe defect detection method based on active stereo omni-directional vision.An omni-directional laser visual sys tem is designed to detect the geometric information of the pipe.The system can obtain the three-dimensional point cloud data in real time.In our system,a crawling robot equipped with the active stereo omni -directional vision sensor (ASODVS) traverses along the pipeline and takes pano ramic image sequences using laser streaks on the inner surface of the pipe.Afterwards,the 3D point cloud data are extracted and the 3D coordinates are calculated.The 3D point cloud data are used to calculate the geometrical feature s,such as minimum diameter,cross sectional area.Based on the geometrical information,the de formation rate of the pipe can be estimated.Since 3D point cloud data are circularly distributed,a triangu lar grid model is used for 3D reconstruction.Experimental results show that our deformation detection and the 3D topography recovery technique are accurate.