Feature point matching method and system based on regional feature expression constraint

Abstract

The invention discloses a feature point matching method and system based on regional feature expression constraint. The method comprises the following steps: S100, partitioning a reference image and an image to be matched; S200, respectively extracting feature points of each block of the reference image and the to-be-matched image; S300, taking each block of the reference image as a query image library, taking each block of the to-be-matched image as a to-be-detected image library, quantifying feature expression of the blocks by using a vocabulary tree, and performing similarity matching on the blocks to find out homonymous blocks; S400, matching the feature points in the same-name blocks to obtain an initial matching set of the feature points; and S500, processing the initial matching set by using a local grid constraint method or a random sampling consistency method, and screening out accurate homonymy point pairs. A high-precision matching result can be obtained for images with large parallax change and geometric deformation, the highest accuracy can reach 100%, and the invention has high application value.

Description

technical field [0001] The invention belongs to the technical field of computer vision image processing, and in particular relates to a feature point matching method and system based on regional feature expression constraints. Background technique [0002] Feature matching is one of the most basic and active research areas in computer vision, and has been widely used in many vision applications, such as 3D reconstruction, object retrieval, etc. Feature selection and extraction are the key to feature-based matching methods. Only by selecting appropriate feature primitives and extraction methods can the accuracy of matching results be guaranteed. The low-level features of images include point features, line features, and surface features. The extraction process of line and surface features is more complicated and must be time-consuming; point features are the most common features in images and are easy to represent and operate. Matching technology has been widely studied and ...

AI Technical Summary

Problems solved by technology

[0005] 1) Due to multiple influences such as weak textures and scale changes, feature point matching errors are prone to occur;

[0006] 2) It is difficult to take into account the correlation of the overall features of the image and the spatial correlation of the point-line features, making the matching results not ideal;

[0007] 3) Mostly rely on deep learning, and have higher requirements for data sets and hardware equipment

Images