Algorithm Study On Comparison Inspection Between Industrial CT Images And CAD Model Based On Edge Extraction

With the increasing of computer-aided design(CAD) technology and the people's requirements of the performance and appearance of industrial products, the design of industrial products are increasingly complex, the difficulty of manufacturing and processing jobs are growing, and how to evaluate the manufacturing quality of the work-pieces is becoming increasingly important. To analyze the work-pieces' manufacture error is an important part of evaluation of their manufacturing process. Industrial computed tomography (CT) technology is an advanced non-destructive testing technique (NDT) which can non-contact and no damage to detect defects and the internal structure of the work-piece, applied to the aerospace, aviation, national defense, military industry, railway transport, machinery manufacturing, and other fields. A method of analyzing the manufacture error based on comparison inspection between industrial CT images and the CAD mode based on edge detection is discussed.In order to achieve the registration of 2D industrial CT image and CAD drawing, we convert the CAD drawing to bitmap file by CAD software, and then extract the sub-pixel edge of the industrial CT image through Facet model. After determined the scaling parameter by minimum bounding rectangle, the rough registration is obtained by determining rotation and translation parameters combined with the mass center and main direction, and then the refined registration is obtained by singular value decomposition and iterative closest point (SVD-ICP) algorithm. Finally, the CAD data is transformed onto the original industrial CT image by the registration parameters, and then the error area will be filled with color to visually display the error.In order to achieve the registration of 3D industrial CT images and CAD model, two feature point sets are needed. The STL format of CAD model is used, and the triangle meshes'vertexes are took as the feature point set after the breakdown of the grid. For industrial CT images, we extract the edge surface as their feature point set. After getting the corresponding slice sequence along three mutually perpendicular directions, the edge extraction of a slice was realized by cellular neural network (CNN) with the adaptive templates. Then the slice edge data were restructured to get the edge volume data along one direction, which would be integrated in three directions by bit or operation to gain the final edge surface. The computer experiment results validated that our algorithm can extract the edge surface of industrial CT volume data more completely, truly and rapidly.If the translation and rotation dislocation between the 3D industrial CT images and CAD model is large, the rate of refined registration will be very slow, and often leads to the refined registration have wrong result because of the local convergence. Therefore, firstly, the rough registration is achieved by the Principal Components Analysis (PCA) with the method of minimal bound box. Then the refined registration between the edged surface data and the CAD model is obtained by SVD-ICP algorithm. The k-d tree is used to improve the calculation speed of searching for the closest point. Finally, the manufacture error is analyzed by compute the distance of each pair of closest points, and the error distribution is visually displayed by color cloud. The experimental results validated that the comparison inspection method is automatic, visualized and high-accuracy. By the improved comparison inspection method of industrial CT images and CAD model, the industrial CT technology can be used to analyze and improve the manufacturing process.