Optimization method of SIFT characteristic matching points based on limit restraint

Abstract

The invention discloses an optimization method of SIFT characteristic matching points based on limit restraint, and aims at overcoming disadvantages of SIFT characteristic matching points. Obtained matching points at present are divided into a high-quality point set V11, a standard point set V12 and a to-be-measured point set V13, the standard point set V12 and to-be-measured point set V13 are updated, the high-quality point set V11, the updated standard point set V12 and the updated to-be-measured point set V13 are combined, a basic camera matrix F is calculated again in an RANSAC method, and exterior points are rejected. Thus, the matching rate is effectively improved, the computational complexity in the budgeting process is low, and computation is rapid.

Description

technical field [0001] The invention belongs to the category of image matching technology in the field of image processing and pattern recognition, and in particular relates to a method for optimizing SIFT feature matching points based on limit constraints. Background technique [0002] Image matching technology is mainly a method to find the same image target through the corresponding relationship, similarity and consistency analysis of image content, features, structure, relationship, texture and gray level. Image matching technology can be divided into: matching technology based on image grayscale, matching technology based on image features, matching technology based on template matching and matching technology based on transform domain. The basic idea of ​​grayscale matching: regard the image as a two-dimensional signal from a statistical point of view, use the statistical correlation method to find the correlation match between the signals, and use the correlation func...

AI Technical Summary

Problems solved by technology

However, no matter which algorithm is used, there is a possibility of errors in the matching process. Therefore, eliminating wrong point pairs during the matching process plays an important role in improving the matching rate.

[0003] After matching the local features extracted by Scale-invariant feature transform (SIFT) in the case of camera translation, the feature points are detected in SIFT, and the feature descriptor is calculated. The Euclidean distance for the feature descriptor In this matching process, in order to ensure the efficiency of matching, the algorithm of K nearest nodes is used for Euclidean distance matching. In order to ensure the correct cross-matching of matching, the matching point pair obtained so far is the correct point in the SIFT algorithm. Yes, due to the limitations of the SIFT feature and the systematic error of the algorithm, some of the matching point pairs obtained are incorrect, resulting in a low matching rate

Images