Research On 3D Reconstruction Of A Single Image Based On SFS Methods

The 3D passive reconstruction technique is an important research direction in the fields of computer vision. It has showed more noticeable status and role in the fields of social productive and life. It has widely application value especially in agriculture, industry, national defense, medicine and space technology field. Due to the Shape Form Shading (SFS) method can reconstruction 3D surface shape only need a single image. The SFS method has development quickly not only in the theoretical research but also practical application in recent years and it has become a research hotspot in 3D reconstruction field.At the beginning, a fully survey about SFS methods is given. Then a systematic summarize is made for all current SFS algorithm in the domestic and foreign, these methods include minimization approaches, propagation approaches, local approaches and linear approaches,we have analyzed their applicability and compared them in terms of convergence rate, uniqueness and existence .Most traditional SFS algorithms need harsh assumptions that don't correspond to the real condition and assume some unsuitable condition so that can not get precise and universality 3D construction result. Based on the research of traditional SFS methods and lighting model, this paper increase the constraint condition of equations, comprehensive equations, Nonlinear least squares principle, surface fitting etc. The paper has reduced the use condition of SFS method, and improving the accuracy of 3D reconstruction. Based on the above-mentioned theoretical research, under the Windows operation system, using Visual C++ 6.0 and MATLAB 7.0 programming environment, an image 3D-reconstruction prototype system is designed and developed.According to experiment result analysis verified this software's 3D reconstruction effect is prominent. Such as rules objects and characters' 3D reconstruction. The results of experiments demonstrate that the algorithms given in this dissertation are feasible, accurate and have good practicability.