Research On 3D Reconstruction Of Objects Based On A Single Image

The 3D prototype reconstruction technique is one of the important research directions of computer vision, and it displays important function in every fields of society production. Now, there are many kinds of 3D prototype reconstruction techniques at home and abroad in computer vision, but shape from shading is one of the key techniques of 3D reconstruction. It reconstructs surface vector and relative height by surface shadeing change of a single image, and makes a basis for 3D prototype reconstruction from 0°to 180°.This paper firstly makes a deep research on current main 3D prototype reconstruction algorithms based on image shading, such as minimization approaches, propagation approaches, local approaches, linear approaches and so on. And then this paper gives a comparison of these algorithm methods, and according to different assumption premise, different limitation and complexity of these four methods, based on these algorithms, this paper gives a new 3D prototype reconstruction method from a single image against calculation complexity of the existed algorithms. It is more simple and applicable. The hardware equipments of this method only include image pick-up system and computer. Its work mode is the uncomplicated principle of optical lens imaging. The core idea of this method starts from the objects surface normal vector and reflect intensity. It takes the object satisfy Lambertian reflentanne Model, so it must exist a point's normal vector direction same to light source normal. Taking the light source direction as Z coordinate to build coordinate and get every point's slant and tilt from curve fomula. Through coordinate transformation, it gets points' space position of objects surface corresponding to every pixel. At last this paper simulate 3D prototype in MATLAB.The experiment result proved that the 3D prototype reconstruction method introduced in this paper can be used to reconstruct objects 3D prototype. The reconstructon by using this algorithm is accurate. The complexity of this algorithm is not very high. Its operation is simple. The environment requirement is decreased. This algorithm has a favoriable application prospect.