View-angle-invariant local region constraint-based slanted image linear feature matching method

Abstract

The invention relates to a view-angle-invariant local region constraint-based slanted image linear feature matching method. The method sequentially comprises the following steps of respectively extracting the linear features of a reference image and a to-be-matched image, and calculating the feature saliency of each linear feature; based on the polynomial mapping function, constructing a view-angle-invariant local region, and calculating the feature region of each linear feature based on the view-angle-invariant local region; for each linear feature, calculating the phase consistency value and the direction in the feature region, and constructing a phase consistency feature descriptor for each linear feature; according to the feature saliency of each linear feature, adopting linear features which are ranked to be the front t% in feature saliency as salient linear features, while adopting all the rest linear features as non-salient linear features; conducting the salient linear feature matching; classifying salient linear features, which are not matched successfully, as non-salient linear features; respectively adopting successfully matched salient linear features in the reference image and in the to-be-matched image as clustering centers, and clustering non-salient linear features into the type of salient linear features; conducting the non exhaustive search method and performing the non-salient linear feature matching.

Description

technical field [0001] The invention relates to the related technical field of image matching in remote sensing image processing, in particular to a straight line feature matching method of an oblique image with constant local area constraints of an angle of view. Background technique [0002] Oblique photogrammetry can simultaneously obtain high-resolution images of the top surface and facade of ground objects, and has been widely used in automatic reconstruction and texture mapping of 3D models of urban buildings, urban planning and monitoring, emergency response, cadastral data verification and update, etc. in the field. Image matching is one of the key scientific issues in oblique photogrammetry data processing. It is the foundation and core of image registration, stitching, 3D reconstruction, target detection and tracking, etc. It has important application value in military and civilian fields. Compared with traditional remote sensing images, there is a large degree of...

AI Technical Summary

Problems solved by technology

However, for oblique images with large viewing angle changes, the matching between line feature intersections is still a difficult problem, which limits the application of such methods in line feature matching of oblique images.

Images