Image data retrieval method and device based on multi-scale vision, equipment and medium

Through non-parametric semantic fusion and text splicing of multi-scale visual features, the accuracy of image data retrieval is improved, and the problems of refinement and accuracy of image data retrieval in the medical and financial fields are solved.

CN120632150APending Publication Date: 2025-09-12PING AN TECH (SHENZHEN) CO LTD
0 Cites 0 Cited by

Patent Information

Application Number
CN202511105015.9
Authority / Receiving Office
CN · China
Patent Type
Applications(China)
Current Assignee / Owner
Filing Date
2025-08-07
Publication Date
2025-09-12

AI Technical Summary

Technical Problem

Existing image data retrieval methods lack accuracy in the fields of healthcare and financial technology, especially in the lack of refinement in understanding lesion levels and key elements, resulting in large errors in image data retrieval.

Method used

A multi-scale visual image data retrieval method is adopted. By obtaining text prompt data of a set of visual example images, multi-scale visual features are extracted, non-parametric semantic fusion is performed, and then they are spliced ​​with text semantic tags. The matching score is calculated to retrieve image data.

Benefits of technology

It improves the granularity, flexibility and accuracy of image retrieval and solves the problem of low accuracy in image data retrieval.

✦ Generated by Eureka AI based on patent content.
Patent Text Reader

Abstract

The invention relates to the technical field of image detection, can be applied to business system platforms of financial science and technology, medical health and the like, and discloses an image data retrieval method, device and equipment based on multi-scale vision and a medium, and the method comprises the following steps: obtaining a vision example image set, and extracting text prompt data of the vision example image set; extracting a visual mark sequence of visual images in the visual example image set; performing non-parametric semantic fusion on the visual mark sequence to obtain a visual fusion mark sequence; performing semantic marking on segmented words in the text prompt data to obtain a text semantic marking sequence, and splicing the visual fusion marking sequence with the text semantic marking sequence to obtain a composite query marking sequence; extracting an image feature mark sequence of an image to be retrieved in the image set, and calculating a matching score of the composite query mark sequence and the image feature mark sequence; and retrieving image data from the image set according to the matching score. According to the invention, the accuracy of image data retrieval can be improved.
Need to check novelty before this filing date? Find Prior Art