A matching algorithm based on global viewpoint difference rectification for framework imagery

Abstract

Matching precision is very important and difficult in accurate photography, remote sensing and 3D reconstruction. After analyzing the advantages and shortcomings of current matching algorithms and the epipolar geometry, a new matching algorithm based on global viewpoint difference rectification for framework imagery is devised. Without the aid of orientation parameters, this algorithm contains three parts, the scale rectification for an overall comfortable of the image, the viewport rectification for compensating the bias between the vertical plane of the primary optical axis and the main plane of the scene, and the consistent transformation of two images with different viewports.. Following, the details of the implementation of our methods are also given. The validation experiments of the matching precision show that, the new algorithm can improve the precision dramatically, which is better than many other good matching algorithms.