Small target detection method and system based on visual attention mechanism

Through the small object detection method of visual attention mechanism, the enhanced cross-space and deformation channel attention modules are used, combined with the multi-branch space attention module, the problem of low detection accuracy of small and medium-sized targets in drone reconnaissance is solved, and higher detection accuracy and background interference suppression are achieved.

CN117274661BActive Publication Date: 2025-09-05ARMY ENG UNIV OF PLA
View PDF 2 Cites 0 Cited by

Patent Information

Application Number
CN202311004051.7
Authority / Receiving Office
CN · China
Patent Type
Patents(China)
Current Assignee / Owner
Filing Date
2023-08-10
Publication Date
2025-09-05
Estimated Expiration
2043-08-10

AI Technical Summary

Technical Problem

During drone reconnaissance, small target feature response is weak and susceptible to complex background noise interference. Existing algorithms are difficult to effectively enhance feature expression and reduce background interference, resulting in low detection accuracy.

Method used

A small object detection method based on visual attention mechanism is adopted, and by enhancing the cross-space attention module and the deformation channel attention module, combined with the multi-branch space attention module, global and local information flow, adaptive channel interaction, and improving feature expression capabilities.

Benefits of technology

It improves the accuracy of small and medium-sized target detection of drone reconnaissance images, enhances attention to target areas, reduces background interference, and improves the learning ability of the detection network.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN117274661B_ABST
    Figure CN117274661B_ABST
Patent Text Reader

Abstract

The present invention discloses a small target detection method and system based on a visual attention mechanism, and relates to the technical field of small target detection. The method comprises extracting multiple levels of initial feature information in the unmanned aerial vehicle reconnaissance image to be tested and inputting the information into the trained enhanced cross-spatial attention module to obtain the first attention map of the current level; inputting the first attention map into the trained deformation channel attention module to obtain the second attention map; inputting the first attention maps of multiple levels into the trained multi-branch spatial attention module for cross-layer integration processing to obtain the third attention map; performing feature fusion on the third attention map and the second attention maps of all levels to obtain a feature fusion map, and then determining the target in the unmanned aerial vehicle reconnaissance image to be tested. The present invention can improve the accuracy of small target detection in unmanned aerial vehicle reconnaissance images by constructing an enhanced cross-spatial attention module, a deformation channel attention module and a multi-branch spatial attention module.
Need to check novelty before this filing date? Find Prior Art

Citation Information

Patent Citations

  • A multi-scale cross attention-based catenary cotter defect target detection method

    CN113658114A

  • Character recognition network model training method, character recognition method, apparatuses, terminal, and computer storage medium therefor

    WO2021115159A1