Multi-scale fusion landmark image retrieval method and system for feature consistency suggestion

Abstract

The invention discloses a feature consistency suggested multi-scale fused landmark image retrieval method and system, and the method comprises the steps: collecting landmark image data, and constructing a landmark retrieval training data set Tr and a test data set Te; constructing a feature consistency suggested multi-scale fused landmark image retrieval network; a total loss function is constructed, a loss value is calculated, a landmark retrieval training data set Tr is used for training a multi-scale landmark image retrieval network, and a multi-scale fused landmark image retrieval model suggested by feature consistency is obtained; and inputting the test data set Te into the multi-scale fusion landmark image retrieval model suggested by feature consistency, and outputting the retrieval result of the landmark image, so that the problem of low retrieval accuracy caused by scale difference under different shooting conditions is solved, the dependence on a large amount of fine-grained label information is reduced, the matching precision of the landmark image is improved, and the retrieval efficiency of the landmark image is improved. And practical application deployment in the field of smart tourism can be realized.

Description

technical field [0001] The invention belongs to the technical field of image retrieval, in particular to a multi-scale fusion landmark image retrieval method and system for feature consistency suggestion. Background technique [0002] With the rapid development of social networking sites, communication and multimedia technology, digital image equipment, etc., the use of digital images has involved all aspects of national defense, military, medical and health, public entertainment and family life. Data such as images and videos are changing with amazing numbers every day. speed growth. In view of these massive images containing rich visual information, how to conveniently, quickly and accurately query and retrieve the images that users need or are interested in from these vast image databases has become a hot research topic in the field of multimedia information retrieval. Landmark image retrieval refers to finding images containing the same landmark building instances from ...

AI Technical Summary

Problems solved by technology

[0005] In order to solve the problems existing in the prior art, the present invention provides a feature consistency suggested multi-scale fusion landmark image retrieval method, which solves the problem of low retrieval accuracy due to scale differences under different shooting conditions, and reduces the need for a large number of details. The reliance on granular label information improves the matching accuracy of landmark images, which is conducive to the realization of practical application deployment in the field of smart tourism

Images