Abstract: Pipe inner surface defects detection is very important to ensure transmission safety and avoid leaking and explosion accidents. Based on the analysis of the geometry characteristic of the 3D points distribution, for the pipe inner side 3D data which are obtained based on circle structured light machine vision technology, a defect detection method about pipe inner side is proposed by comparing the normal vector angle between adjacent points which are on the same circumference. The point normal vector is calculated by the weighted average of adjacent triangles normal vectors. As the 3D points distribute circumferentially, the adjacent points are adopted to do fast triangulation. By using above method, we conducted defects detection and 3D reconstruction of measured and simulated 3D points inside the pipe, respectively. Results show that the method can detect and identify pipe defects of less than 0.1 mm racial variation with 0.081 mm accuracy.