A Feature Matching Algorithm Based on Topological Structure of Feature Points

Abstract

The invention discloses a feature matching algorithm based on the feature point topology structure. First, the two-way optimal matching point pair is calculated through the cosine similarity of the feature point descriptor vector, and then the error is eliminated by calculating the topology structure of the two-way optimal matching point pair. Matched point pairs, and then use the remaining matching point pairs to calculate the homography matrix, and then map all the feature points in the registered image, eliminate the point pairs whose coordinates do not match after mapping, and obtain the final matching point pair, according to each matching Score the logarithm of points, and keep the corresponding matching score; compare all the matching scores, keep the largest matching score as the matching score between the registered image and the input image, and the matching ends. The present invention can calculate a more accurate and stable homography matrix, so that the effect of matching points is better, and it can ensure that the matching results obtained when a pair of images are matched multiple times are the same, and it is avoided that an image that could have been matched in a certain time When it is judged that the image does not match during matching.

Description

technical field [0001] The invention relates to a safety bolt device, in particular to a feature matching algorithm based on a feature point topology, and belongs to the technical field of safety bolt devices. Background technique [0002] At present, in the field of fingerprint recognition, there is a common problem that the matching accuracy of feature points is low. In a fingerprint image, there may be multiple local regions that are very similar, resulting in very similar descriptor vectors of fingerprint feature points, which becomes difficult in the feature matching stage. When encountering low-quality fingerprint images including noise and distortion, the effect will be worse and the matching accuracy will be lower. [0003] The existing feature point matching schemes in the industry mainly calculate and match from the Hamming distance of the descriptor vector, and remember that the two images to be matched are the registration image and the input image. The specifi...

AI Technical Summary

Problems solved by technology

Firstly, the number of iterations cannot be controlled. Secondly, because it is randomly selected, the points for calculating the homography matrix may be different each time, resulting in different matching results each time.

Images