An Improved Fundamental Matrix Estimation Method for View Geometry
In the multi-view geometry,the estimation of fundamental matrix needs to use accurate corresponding points,but the image noise or light changes will lead to wrong matching and affect the accuracy of the estimation of fundamental matrix.In order to solve this problem,an iterative method based on epipolar constraint gradient is proposed to estimate the fundamental matrix.This method can quickly eliminate the wrong matching points in the matching process of two views in relatively little iteration.Then by adding corresponding point correlation constraints,the optimal set of corresponding points can be obtained to estimate the fundamen-tal matrix.Experiments on different scene figures show that compared with other methods,the proposed method can reduce the error of fundamental matrix by at least 20%.
fundamental matrix estimationmulti-view geometryepipolar constraintcomputer vision
国家科技期刊平台
NSTL
万方数据
