Dissertation > Industrial Technology > Automation technology,computer technology > Computing technology,computer technology > Computer applications > Information processing (information processing) > Pattern Recognition and devices > Image recognition device

The Research and Implement on Camera Calibration Technology Based on Trifocal Tensor

Author HanWeiWei
Tutor LiuFeng
School Nanjing University of Posts and Telecommunications
Course Signal and Information Processing
Keywords Fundamental matrix Trifocal Tensor Levenberg-Marquardt algorithm Genetic Algorithms Camera Calibration
CLC TP391.41
Type Master's thesis
Year 2012
Downloads 40
Quotes 0
Download Dissertation

In recent years, with the rapid development of computer technology and digital products, computer vision research is being more and more attention of the people. As one of the main research direction of computer vision, camera calibration is a two-dimensional digital image restoration and reconstruction of the three-dimensional characteristics of objects and a necessary step. Existing camera calibration techniques include traditional methods, based on the method of active vision and camera self-calibration method. Traditional methods need to shape geometry information known calibration object as a reference, it is difficult to apply in practice; method based on active vision need to use the camera to do a particular movement particularity to achieve calibration, high accuracy requirements of the experimental system; based self-calibration method of calibration scene and calibration of the instrument less demanding, but often poor robustness, accuracy is not high. To this end, we study trifocal tensor estimation algorithm and a trifocal tensor-based camera self-calibration method, the concrete work is as follows: First, the basic theory of computer vision to do a simple generalization. Describes the geometric model of camera imaging geometry and multi-view camera imaging geometry model including linear and nonlinear models, linear model is relatively simple, and the non-linear model can be used to simulate and compensate for camera various aberrations, but often high computational complexity and relatively difficult to implement. Second, the study of the fundamental matrix of epipolar geometry estimation algorithm. Basic matrix represents a specific scene for the transfer between two view epipolar geometry relations, initial match points to get through the corner matching algorithm based on gray correlation based on RANSAC algorithm estimates fundamental matrix, at the same time to get a precise matching points between the two images. Again, in a detailed study the existing multiple trifocal tensor estimation algorithm based on the proposed estimation algorithm based on genetic algorithm and LM algorithm trifocal tensor. Existing trifocal tensor estimation algorithms including the direct linear solution method, the iterative algorithm and RANSAC algorithm. Calculation of the direct linear algorithm is relatively simple, but not easy to obtain the optimal solution, iterative algorithm complexity depends on the initial selection RANSAC algorithm robustness, but the selection of the threshold is not easy to determine. The proposed trifocal tensor estimation algorithm makes full use of the advantages of genetic algorithm global search ability and the LM algorithm iteration decline fast, the experimental results show that this method increases the computational complexity, but more precise results. Finally, based on the the trifocal tensor camera calibration method, this calibration method belongs to the category of absolute dual quadric surface-based camera self-calibration method and general self-calibration method trifocal tensor can get the same reference to the three camera projection matrix in the coordinate system, and finally use iterative optimization of the initial results of the non-linear equations, linear processing, to obtain a more accurate camera parameters. The experimental results show that the self-calibration method is correct and effective.

Related Dissertations
More Dissertations