Three-dimensional defects detection of high-voltage cable joint based on iterative residual fitting

Abstract Aiming at the three-dimensional (3D) defects detection of high-voltage cables joint, a method based on iterative residual fitting is proposed. The 3D point cloud data was obtained by reconstructing the laser line projected onto the surface of the high-voltage cable joint. In order to remove noise generated by reconstruction which is located above or below the curve, a simple and effective strategy that the noise can be filtered by the gradient change in the neighborhood is adopted. Then a continuous iterative residual fitting method is proposed to effectively reduce the interference of the defect area on the curve fitting accuracy, thereby obtaining the accurate equation expression of the cable surface curve. By comparing the residual value of the high-voltage cable surface points relative to the fitting curve, the points of the defect area are marked. Finally 3D defect area is extracted through segmentation and clustering. Experiments have verified the effectiveness of the proposed method on artificial standard size defects and real manual ground defects. The absolute error of the depth (height) of artificial standard size defects is less than 0.05 mm. Compared with the traditional cylindrical fitting method, it can be seen that the recognition rate of the proposed method for manually ground defects of high-voltage cable joint is much higher than that of cylinder fitting method.