Image Feature Point Matching Method Preserving Spatial Structure

Abstract

The invention discloses an image characteristic point matching method for maintaining a space structure, comprising steps of extracting and symbolizing the characteristic points of an image I and an image II to be matched, constructing a description sub-distance matrix of the characteristic points of the image I and the image II, constructing a coordinate distance matrix of the characteristic points of the image I and the image II, normalizing the coordinate distance matrix of the characteristic points of the image I and the image II, using the normalized coordinate distance matrix and the description substructure distance of the optimal matching characteristic points of the image I and the image II to perform iteration and optimization on the description sub-distance matrix, and determining and outputting the matching degree according to the description sub-distance matrix element change and iteration frequency after the iteration and optimization. The invention improves the accuracy of the characteristic point matching because the space structure matching is integrated in the characteristic point matching.

Description

technical field [0001] The invention relates to the field of pattern recognition and computer vision, in particular to an image feature point matching method for maintaining spatial structure. Background technique [0002] In computer vision applications, the common method of image matching is to extract feature points from each image first, and use descriptors to represent each feature point, and then compare the similarity of descriptors to match feature points of different images. [0003] The output of image feature point extraction and characterization is the position and descriptor of image feature points, which is the previous step of image feature point matching. SIFT (Scale Invariant Feature Transform) is a classic image feature point extraction and representation method. The main steps include constructing scale space, detecting and filtering Gaussian difference extreme points, assigning direction values ​​to generate descriptors, etc. [0004] Image feature point...

AI Technical Summary

Problems solved by technology

[0005] In order to solve the problem that the current image matching method does not consider the positional relationship of feature points and easily lead to errors, the present invention provides an image feature point matching method that maintains the spatial structure, which incorporates the matching of the spatial structure in the feature point matching process, Improved accuracy of feature point matching