Fast Matching Method of Color Invariant Synthetic Features Constrained by Spherical Panoramic Epipolar Lines

To address the problem of low automation of feature matching in stereo panoramic model measurement applications, a color invariant synthetic feature matching method based on epipolar geometric constraints is proposed. Firstly, the method limits the search range of panoramic image matching from two-dimensional to one-dimensional band buffer by epipolar line constraint. Secondly, the method uses the color invariant correlation coefficient to further determine the fine search range. Finally, the method establishes a kind of matching synthetic measure model to complete panoramic feature matching based on color invariant feature and the rotation invariant local binary patterns. After taking panoramic images in the field and feature matching nearly 200 groups of points, the matching accuracy of the method proposed in this paper is improved by nearly 10% compared with the gray-scale matching method, SIFT, SURF and CSIFT. The matching error of the proposed method can be decrease greatly, and the reliability and precision of spherical panorama image matching is effectively improved, which can lay the foundation for secondary sampling, measurement and depth map generation of panoramic measurement model.