3D Reconstruction Of Pipeline Welds And Identification Of Surface Defects Based On Image Sequences

Modern industrial production is becoming more and more intelligent,and the proportion of welding in the manufacturing industry is also very large,in order to ensure a high rate of qualified pipe welds during welding operations,it is particularly important to achieve intelligent identification of pipeline weld defects,this paper designs a single camera-based multi-location pipeline weld 3D surface defect identification system,3D reconstruction of the pipeline weld,to identify the pipeline weld surface defects and find the defect This paper designs a single camera-based multi-location pipeline weld 3D surface defect identification system,which performs3D reconstruction of the pipeline weld,identifies the pipeline weld surface defects and finds the location of the defects,and achieves automated identification of weld surface defects.This paper describes the camera imaging principle monocular vision system construction and camera calibration.The imaging principle specifically includes:projective geometry,coordinate transformation and pinhole model;the construction of the monocular vision system includes:the design of the monocular vision system,camera selection and camera distortion model;the calibration of the camera includes:the principle of calibration and calibration experiments and results.3D reconstruction of pipe welds.Introduce the principles and methods used in each step of 3D reconstruction,including:the principles of sparse point cloud reconstruction,dense point cloud reconstruction and Poisson surface reconstruction;use the multi-view pipeline weld surface defect recognition experimental platform for multi-view image acquisition,and carry out reconstruction work on the acquired images according to the process of Sf M,PMVS2 and Poisson surface reconstruction to obtain a complete 3D model of the pipeline weld.Pipeline weld surface defect identification.To complete the pre-processing and defect identification of the 3D point cloud,firstly,noise removal and smooth processing are performed on the collected dense point cloud of the pipe weld,and secondly,the magnitude of the principal curvature k₁ and k₂,Gaussian curvature K and average curvature H are calculated for each point in the point cloud.According to the experimental data,we know that the point clouds with the value of K between-3e-06 and 3e-06 are qualified,and those outside this value are defective;and K=0,H=0,K>0,H>0 and K>0,H<0 are also defective point clouds.The software for 3D visual recognition of pipeline welds is designed.The 3D surface defect recognition software for pipeline welds designed in this project is introduced in detail.The specific functions include:image acquisition,camera calibration,point cloud reconstruction and defect recognition;and the actual effect of each function is verified through experiments:the actual sampling data and recognition results are calculated and compared to obtain the accuracy rate of defect recognition.The experimental results show that the software and hardware system and related algorithms designed in this project can realize the recognition of surface defects of pipe weld tumor,weld penetration,porosity,and weld deviation,and the recognition accuracy rate reaches 96.23%.