Unmanned aerial vehicle aerial small target detection method and system based on frequency domain-wavelet fusion

By using the frequency domain-wavelet fusion method of LOWTC and CFSA modules, the problems of low detection accuracy and noise interference of small targets in UAV aerial images are solved, and high-precision and efficient UAV small target detection is achieved.

CN122289970APending Publication Date: 2026-06-26SHANDONG UNIV
View PDF 0 Cites 0 Cited by

Patent Information

Application Number
CN202610258493.1
Authority / Receiving Office
CN · China
Patent Type
Applications(China)
Current Assignee / Owner
Filing Date
2026-03-04
Publication Date
2026-06-26

AI Technical Summary

Technical Problem

Small target detection accuracy in drone aerial images is low and susceptible to noise interference. Existing technologies such as YOLOv8 have low average accuracy on drone datasets, and Wavelet convolution lacks a frequency domain attention collaboration mechanism.

Method used

The low-frequency enhanced wavelet transform convolution module LOWTC and the Fourier frequency domain self-attention convolution module CFSA are adopted. The receptive field is expanded through second-level wavelet decomposition, noise is suppressed and global dependence is enhanced, and a new small target detection head is added to detect targets smaller than a preset pixel.

Benefits of technology

It significantly improves the detection accuracy of small targets in UAV aerial images, increases APs by 5.5 percentage points, reduces the number of parameters, and is suitable for embedded real-time detection scenarios.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN122289970A_ABST
    Figure CN122289970A_ABST
Patent Text Reader

Abstract

This invention belongs to the field of computer vision and deep learning technology, and provides a method and system for small target detection in UAV aerial photography based on frequency domain-wavelet fusion. A small target detection head branch is added after the second layer output of the backbone network: after downsampling to a preset pixel resolution, it is adaptively weighted and fused with the deep features output by the detection head through a structure. A decoupled head structure is used to output classification prediction, bounding box regression prediction, and confidence prediction respectively, for detecting small targets with a resolution smaller than the preset pixels. The feature map output by the backbone network is input into a LOWTC-based feature extraction module, which performs two-level wavelet decomposition, recursively convolving the low-frequency subband to capture long-distance context, and preserving edge details in the high-frequency subband. After reconstruction by inverse wavelet transform, it is fused with the deep semantic feature structure at multiple scales. This solves the technical problems of low detection accuracy and susceptibility to noise interference in small target detection in UAV aerial images.
Need to check novelty before this filing date? Find Prior Art