Image retrieval method based on graph online hashing model

By using a supervised online hashing method based on graph neural networks, image features are extracted and a graph online hashing model is constructed, which solves the problem of low efficiency of shallow models in existing technologies and realizes efficient online retrieval of image data.

CN117112827BActive Publication Date: 2026-06-26OCEAN UNIV OF CHINA
View PDF 2 Cites 0 Cited by

Patent Information

Authority / Receiving Office
CN · China
Patent Type
Patents(China)
Current Assignee / Owner
OCEAN UNIV OF CHINA
Filing Date
2023-08-28
Publication Date
2026-06-26

AI Technical Summary

Technical Problem

Most existing online hashing methods are based on shallow models, which cannot effectively handle real-time retrieval of large-scale dynamic image data, and the high cost of updating hash functions leads to low efficiency.

Method used

A supervised online hashing method based on graph neural networks is adopted. Image features are extracted through visual graph neural networks, and a graph online hashing model is constructed. The objective functions of similarity preservation loss, classification loss and knowledge preservation loss are used to learn to generate compact hash codes.

Benefits of technology

Deep online hashing was implemented, which improved the accuracy and flexibility of online image data retrieval, effectively processed streaming image data, and improved retrieval performance.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure QLYQS_2
    Figure QLYQS_2
  • Figure QLYQS_18
    Figure QLYQS_18
  • Figure QLYQS_24
    Figure QLYQS_24
Patent Text Reader

Abstract

The application discloses an image retrieval method based on a graph online hash model and belongs to the technical field of image retrieval.The application divides each image into multiple small blocks as nodes, adopts a visual graph neural network model to extract image features, and further inputs each image as a node into a graph online hash model, learns an aggregated representation function of adjacent nodes, constructs a GNN hash model to generate a feature vector of a target node, and designs a target function composed of a similarity preserving loss, a classification loss and a knowledge preserving loss to learn a compact hash code with strong semantic representation capability for image retrieval.The application utilizes the advantage that a deep network can mine more semantic features, and greatly improves online retrieval accuracy for image stream data.
Need to check novelty before this filing date? Find Prior Art