Stereovison is a method which gets three dimensional geometric information of objects from images.It embodies theories and schemes of computer vision, images processing as well as computer graphics and has been the cutting-edge research problem. Passive three dimensional method which does not need artificial treatment in scene,can acquire the space information of scene and abtain the goal of three dimensional reconstruction,and it has a promising future.This thesis primarily works on algorithms of three dimensional reconstruction from images under projective geometry meaning of stereovison.This method only needs weak calibration for cameras and can achieve three dimensional reconstruction. This thesis finishs the following work:1.Illustrates the theories of three dimentional reconstruction under projective geometry meaning,and confirms methods and processes of three dimensional reconstruction.2.Using Harris and SUSAN arithmetics,extracts the feature points,and matchs these points through correlation ,then disambiguates matches through relaxation.3.Summarizes all arithmetics of calculating fundamental matrix and compares these arithmetics,ascertain M-Estimators has a better result.4.Decomposes the fundamental matrix into two projetive matrix under projective geometry meaning ,extracts and calculates primitives'coordinates in space.5.Using MATLAB,Visual C++ and OpenGL,reconstructs three dimensional model and validate these methods are feasible. |