Study And Implementon On Comparison Inspection Between Industrial CT Model And CAD Model

With the demand for the complexity of the industrial products has increased, thedifficulty of detection is getting higher and higher for the products, which are after themanufacturing process by computer-aided design (Computer Aided Design, CAD).Inmany test methods, industrial CT(Computed Tomography) testing because of its abilityto non-destructive testing for internal structure and the properties and materials ofworkpieces, become an excellent means of detection. During the detection to theworkpiece, the key step is the alignment between CT slice reconstruction model withthe original CAD model in order to get the correct detection results. In many algorithmsabout the field of3D registration, there are many studies based on3D point cloud fromthe laser scanning, while a few studies are based on points from industrial CT, and thestudies are poorly targeted, and there is a large distance from the practical stage.Based on the processes of Comparison Inspection between industrial CT scanreconstruction model and the CAD model, and characteristics of the objects that arescanned by industrial CT, in this paper the initial registration method which is able toadapt to the complexity of the workpiece.On the analysis of the various initialregistration methods, and on the basis of the classical principal componentanalysis(PCA) method as the prototype, by introducing the relationship between theposition of point clouds’ center and centroid, a modified method is prompted, to judgethe registration if there is reverse happen after the principal component analysis method,then it reverses registration model, and get the right initial registration in a relativelyshort time.In order to achieve precise alignment between the two models in the initialregistration stage, the closest points search method and the rotation matrix method inthe standard iterative closest point algorithm (ICP)are discussed. Through theintroduction of k-nearest neighbors principle, the new method establishes aneighboring point quick search tree, to improve search speed. By using singular valuedecomposition(SVD) method, the new method solves question of the rotation matrix.compared to the standard ICP algorithm and the improved methods’ solution speed,thenew method has higher speed and the same calculation accuracy range.Finally, a industrial CT scan reconstruction model and the CAD model comparisonsoftware is developed, based on OpenInventor,which is a three-dimensional graphics package.The initial registration and precise registration methods are insert into thesoftware, and it provides the related support functions of registration such as the displayfunction, the resulting output function.