3D Reconstruction Techniques Based On Image Sequences

Image-based3D reconstruction is a very active research topic in the field of computer vision, this method has low cost, easy to use and direct access to the color information of the surface characteristics and to resolve the current number of three-dimensional model is far from meeting the practical application of The demand for this situation.3D reconstruction of the Qin Terracotta Warriors and Horses debris, the use of a camera image sequences combined with computer vision and graphics image processing technology, the key to3D reconstruction algorithm based on multi-image sequence. The main innovation is as follows:(1) Conversion of the fundamental matrix solution:In this paper, global optimization strategy, the use of nonlinear methods to estimate the fundamental matrix. First of all, the fundamental matrix of the inherent nature of linear algebra calculation skills, with a minimum of parameters to express the fundamental matrix. Second, the polynomial function to establish the mathematical model of fundamental matrix estimation, model optimization problem is transformed into a polynomial of the minimum problem. Finally, the use of more reckless sexual RANSAC algorithm is selected to meet the minimum distance error of the fundamental matrix as the final solution, in order to improve the robustness of the fundamental matrix solution.(2) Non-linear camera calibration improvements:Through in-depth study and analysis of existing non-linear camera calibration method, the original radial distortion on the basis of lens distortion and cut to the distortion, combined with manualcalibration method to complete the calibration. The experiments show that the method can meet the calibration requirements of this study.(3) Projection matrix series solution method:Derive the fundamental matrix and projection matrix mathematics iterative relationship, the series of solving a sequence of image projection matrix in the projective space, combined with the camera calibration, the projective space of the projection matrix transformations to measure the space. Experiments show that because the basis matrix with higher robustness of the proposed method to calculate the camera projection matrix is robust.(4)Two-dimensional back projection of the image to minimize the error treatment methods:The use of three two-dimensional image visible more and be able to provide a stable geometric relation constraints, this article uses the nature of the recovery of3D points in a metric space with the projective backprojection image point iteration space to minimize the error handling. Minimize anti-projection error processing, the constant correction errors, and optimize.