Three-dimensional Reconstruction Technology And Its Application

Modeling based on images is defined how a 3D surface model can be obtained from a sequence of images taken with off-the-shelf consumer cameras. The user acquires the images by freely moving the camera around the object .Neither the camera motion nor the camera settings have to be known .The obtained 3D model is a scaled version of the original object , and the surface texture is obtained from the image sequence as well.In this text ,the first steps of modeling from images are discussed. It consist of computation of fundamental matrix in two views and of trifocal tensor in three views . we consider several algorithms of computation of fundamental matrix and condition when they are used. Computation of fundamental matrix and trifocal tensor are unified into one frmae. In experience ,It is tested that MAPSAC algorithm is very robust ,even if there are big ratio outliers. Because MAPSAC algorithm is based RANSAC. Once fundamental matrices are computed then the camera projective matrices may be recovered in every views and can used to estimate the structure using generic triangulation algorithm. Using projective matrices of every views, each 3D points are projectived again and projection error is computed. Using those datum, fundamental matrix is tested. Is it 'good' or 'bad'?By analyzing characteristics of methods of computing fundamental matrix. A new method is presented for robustly estimating fundamental matrix . There are three contributions , the first is discriminate data set into inliers or outliers by minimizing reprojection error . the second contribution is computation of one order approximation of reprojection error . Avoiding local minimum , the third contribution is adoption of LM algorithm in last steps of RANSAC algorithm .A very good estimation of fundamental matrix is obtained and reprojection erroris more smaller . Experimental results on synthetic and real images demonstrated that the new algorithm is valid and robust.