Three-dimensional Reconstruction Of Metal Fracture

Fracture surface analysis is widely used in fracture failure analysis. People attach important to quantitative analysis of fracture surface which is the further development based on qualitative analysis in recent years. The measure of metal fracture is performed mostly using the Scanning Electron Microscope (SEM). Apparently, this qualitative description based on 2-D image is lack of accuracy. In this paper, the surface shape of metal fracture is recovered by 3D reconstruction technology based on a single metal fracture SEM image. 3D feature of metal fracture can be analyzed quantitatively. It is very useful for the further study.Shape from shading (SFS) is one of the main developments of the computer vision that deals with reconstructing the 3-D shape of an object from single two-dimensional (2-D) image intensity. The conventional approaches are under the restrictive condition of the Lambertian model, in which the light source direction must be given or reasonably estimated. Apparently, the reflectance of real secondary electron SEM does not follow these restrictions. The conventional algorithm is impractical for metal fracture SEM image.As to the problems above, the paper takes quantitative analysis of fracture surface as applying background. We combine the 3D reconstruction and relative problems in the computer visual research area to make systematic research, to apply and improve the available 3D reconstruction method based on analyzing and summarizing the advantages and disadvantages of all the available methods. More important is that this paper puts forward two 3D reconstruction models of metal fracture SEM image. According to the working principle of Scanning Electron Microscope and the fractal characteristic of metal fracture, a reflection model based on finite-element and a new neural-based reflectance model of the SEM image have been established. The smoothness constraint has been replaced by fractal constraint and fractal interpolation has been applied to finite-element model in order to overcome the disadvantage of linear interpolation. The experiment results show that the detail features of metal fracture surface are reconstructed better than conventional SFS algorithm and the brightness errors are obviously decreased. The proposed two algorithms in this paper are proved to be with high efficiency and good precision.In this paper, the level of material surface analysis is improved by studying surface characteristic of the fracture in 3D space. It states position and function of fracture surface analysis in fracture failure analysis.