Research On 3d Reconstruction And Defect Detection Of The Pipeline

The pipeline is frequently used as a non-substitutable device in long distance transport, which has great influence on our daily life and industrial production. However, various problems occur since the pipeline service duration increases, causing troubles without attention. The detection becomes one key point in the pipeline maintenance. We need enough data of the pipeline and process these data to achieve useful informations. This is of great importance in society and economy.In order to detect the pipeline defects, this paper focus on the 3d registration and surface reconstruction. This paper investgates the current state of the pipeline defect detection and 3d reconstruction methods, showing the significance of 3d reconstruction in pipeline defect detection. This paper presents a deep analysis of the reconstruction methods, points out the weakness and analyzes the covariance sampling method which can keep the geometric stability in regular goemetric models, pays attention to the density variable poisson surface reconstruction method. Through the analysis of the current 3d reconstruction methods, as well as the geometric characteristics of the pipeline, this paper designs the 3d registration methods, including coarse registration estimation, coarse registration optimization and fine registration with corresponding points estimated by covariance matrix. This paper goes throuth the pipeline defect detection techniques and proposes a variable scale method in nearest neighbor search in point cloud data. This variable scale method solves the difficulty between the details reserving and the re liability of the nearest neighbor search. The 3d distance data and image informations are obtained through the wheel robot platform equipped with CCTV system and laser range finder. The designed reconstruction method is implemented by programming language and the primitive data are processed to obtain the specific goals. The effectiveness of the designed method is proved and the analysis of the experiment results is in detail.Through the experiment results, the method presented in this paper for 3d registration limits the axis transformation registration error at about 3%, the circumferential direction twist error at about 3°. The pipeline defects’ dimention and position detection error are about 3%. The method presented in this paper can result in efficient 3d reconstruction and defect detection.