Shape Analysis Based On Laplacian Deformation

Surface defects on sheet metal panels, which are commonly assembled for outer cover shell of industry products, cause unwanted visual flaws deviating form the intended appearance. Surface defect inspection is involved to detect unqualified panels or to be helpful to improve the producing technics. With modern CAE (Computer-Aided Engineering) technology, the application of punch forming emulation systems which simulate the real producing process of sheet metal panels are widely used for the sake of decreasing the tryout stage and of cutting the developing cost. The output panel models represented by finite element mesh from the emulation systems give high fidelity to the real shape of the stamped panels, simulating spring back and having realistic surface detail. It is desirable to conduct the surface shape analysis directly after stamping simulation to visual or detect the latent defects on the virtual panel. In this dissertation we present a novel shape analysis technology, which measures defects through comparing shape details of the defect scale between the defective and the reference panel. This new technology also provides a convenient measurement for automatic defect grading and evaluation.The method we describe here is deferent from the previous work, for it not only analyzes the geometric information alone but also takes fully the advantage of the reference shape. It needs no auxiliaries like structure light and reflection system, but is just based on the geometric deforming techs. Using Laplacian deformation and spectral decomposition, we can generate a couple of new defective and reference models with corresponding vertices and the same mesh topology from the originals. Defect information then is deduced from normal deviations of their corresponding vertex positions.