The Research Of 3D-Reconstruction From Sem Image Of Clay Soil

The 3D Reconstruction is a technology that to build up the object's 3D digital model from the information that directly or indirectly about the object. And the digital models which included the body's spatial geometric info can be read by computers so as to do some more complex analyze or process about the object by the computers. Just because the model can be processed by the computer,3D Reconstruction technology has been used in various realms, including machine, geography, simulation, film and games etc.3D Reconstruction is based on the info about the object body, so what info or what formatted info about them decided the algorithm or the processing. There is not a uniform method or procedure that can be defined to the reconstruction.SFS(Shape from Shading) is one of the 3D Reconstruction technology, which is intend to reshape the object's 3D surface from the one or more photos of the body. SFS may be look up as the reverse-procedure of taking photo, that says while taking photos the camera translates the real object's image into film, but SFS is to translate the 2D image into 3D virtual object. Generally SFS needs only one picture to reconstructive the 3D model, so this technology is widely used and developed in many realms.In the civil field, the micro-structure is one of the import factors which are to decide the force characteristic of soil. And it is a common method to use the Scan Electron Magnifier (SEM) to study the clay soil's micro-structure. Engineers can get the image (says the SEM Image) of soil's micro-structure with SEM. The SEM image is a gray color picture and the gray info (rank from 0 to 255) shows the much of 3D info about the soil, including terrains and shading. From the SEM image, the engineer can see the micro-structures and do some qualitative or even quantitative studies. When doing quantitative studies, it is a classic method to translate the gray image to white-black image by a point valve. The pixels whose gray color is greater than the valve are set to white and regarded as the soil granule, the others are set to black and regarded as the hole between the granules. Now geo-shape of soil can be got from the white-black pictures using the image-proc tech., and some quantitative studies about the soil can be done by the computer through the geo-shape.The classic procedure looks the gray info as the terrain of soil sample, but only use one or zero to distinguish the soil or the hole. That says the other gray info is ignored, and the surface of soil is not continuous changed but in fact it is do. If we can use all the gray info properly which including the 3D surface info to reconstructive the soil's structure, and build up a digital model of soil's micro-structure, we can get more quantitative studies and get more accurate result. How to do it?This article will try to resolve this problem. One of the methods is to look the gray info as the terrain of soil surface, and then a matrix of the terrain value about the soil surface can be got from the SEM image. This proceeding is called terrain-method in this paper. The other method is using the shape info that included the SEM, and a normal vector matrix about the surface will be build-up. And then using the vector matrix, the 3D model of soil's structure will be reshaped. This is called normal-vector method in this paper.