Image matching method and device and related equipment

Abstract

The invention discloses an image matching method. The method comprises the steps of obtaining a to-be-detected image and a sample image; processing the to-be-detected image and the sample image by using a preset network model to obtain a target feature vector and a sample feature vector; performing binarization processing on the target feature vector and the sample feature vector to obtain a target binary feature and a sample binary feature; performing distance measurement on the target binary feature and the sample binary feature to obtain a feature distance; determining an image matching result according to the feature distance. The image matching method is simple and fast in implementation process, and can effectively improve the image matching efficiency. The invention further discloses an image matching device and equipment and a computer readable storage medium, which all have the above beneficial effects.

Description

technical field [0001] The present application relates to the technical field of image processing, in particular to an image matching method, and also relates to an image matching device, equipment, and computer-readable storage medium. Background technique [0002] In the field of computer vision, the application of deep learning technology for feature learning has been widely used. By building an appropriate model framework and selecting an appropriate optimization cost function, the input image or video can be reasoned through the network to obtain the corresponding representational features; further, through an appropriate distance measure, these features can be matched, retrieved, etc. And it is widely used in image / video matching, image / video retrieval, video tracking, image verification and other fields. [0003] Among them, in the existing local image matching problem, matching an image no longer uses a single characteristic feature, but requires a large number of l...

AI Technical Summary

Problems solved by technology

[0003] Among them, in the existing local image matching problem, matching an image no longer uses a single characteristic feature, but requires a large number of local image representation features. Therefore, the matching calculation between images will no longer be a single The distance measure of features, but the pairwise distance calculation of a large number of local features, takes a lot of computing time

Some design and training methods based on binary networks have been proposed in the prior art, but these methods focus on the acceleration of model reasoning. In the process, the model needs to be redesigned and trained, and the computational efficiency cannot be effectively improved. , so it is difficult to guarantee the efficiency of image matching; moreover, this implementation method does not have a mature binary reasoning framework for mobile applications, and the implementation process is relatively difficult

Images