Target Image Retrieval Method and System

Abstract

The invention discloses a target image retrieval method and system, wherein the method includes: extracting sample feature points of each image sample in an image sample library; filtering out redundant sample feature points in each image sample; Construct public feature point set and private feature point set of sample feature points; extract the feature points to be processed of the image to be processed; combine the feature points to be processed with the sample features in the public feature point set and the private feature point set Points are matched to obtain the matching score of each image sample; the image sample with the highest matching score is selected as the target image. The present invention simplifies the image sample library by compressing the redundancy of sample feature points, which can significantly speed up the matching speed while reducing the number of matching times, thereby improving the matching accuracy, so that the entire retrieval process is more efficient and effective than the existing technology. With obvious improvement.

Description

technical field [0001] The invention relates to the field of image processing, in particular to a target image retrieval method and system. Background technique [0002] Feature point matching algorithm is one of the key technologies to realize image stitching, image fusion, and target recognition and tracking. How to improve the efficiency and accuracy of image matching has always been a hot spot in the field of image processing and machine vision research. [0003] The existing target retrieval method based on feature point matching algorithm mainly includes two parts: feature point extraction and feature point matching. Image feature point extraction refers to detecting representative key points in the image and obtaining feature descriptors of key points. Usually, SIFT, SURF and ORB algorithms are used to extract feature points and feature description operators of images; feature point matching It is usually based on the feature description operator. The above feature d...

AI Technical Summary

Problems solved by technology

[0006] 1) Excessive extraction of feature points and redundancy, resulting in at least one round of comparison is required for each additional feature point, and it contains a large number of unnecessary comparison processes, which seriously reduces the matching efficiency

[0007] 2) Comparing the matching method of the distance between the nearest neighbor and the next nearest neighbor, although the feature points without matching relationship caused by image occlusion and background confusion can be excluded, the image samples in the image sample library also have similar redundant feature points, which leads to the nearest If the distance between the neighbor distance and the second nearest neighbor distance sample feature points becomes smaller, the matching will cause misjudgment, and the reliable feature points will be filtered out by mistake.

Images