Lightweight remote sensing small target detection method based on edge enhancement and multi-dimensional attention

A lightweight remote sensing small target detection method based on edge enhancement and multidimensional attention solves the problems of background noise and multi-scale feature representation in remote sensing images, and achieves efficient small target detection.

CN122090170APending Publication Date: 2026-05-26SHENYANG UNIVERSITY OF TECHNOLOGY
0 Cites 1 Cited by

Patent Information

Application Number
CN202610469591.X
Authority / Receiving Office
CN · China
Patent Type
Applications(China)
Current Assignee / Owner
Filing Date
2026-04-10
Publication Date
2026-05-26

AI Technical Summary

Technical Problem

Existing methods for detecting small targets in remote sensing images struggle to integrate redundant background noise in complex contexts and fail to adequately represent multi-scale features when the target scale changes drastically, resulting in low detection accuracy.

Method used

A lightweight remote sensing small target detection method based on edge enhancement and multidimensional attention is adopted. The frequency domain guided gated edge enhancement module (FGEM) is used to suppress noise and enhance texture. The lightweight multidimensional attention module (LMDA) is combined to express multi-scale features. The phantom guided reparameterization module (PGM) is used for feature extraction to achieve detail enhancement and context modeling.

Benefits of technology

It significantly improves the accuracy and robustness of small target detection in remote sensing images, reduces computational complexity, adapts to complex backgrounds and target scale changes, and maintains the lightweight characteristics of the model.

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

Abstract

The invention discloses a lightweight remote sensing small target detection method based on edge enhancement and multi-dimensional attention, and the method comprises the steps: firstly, carrying out the multi-scale feature extraction of a remote sensing image based on stack phantom guidance re-parameterization, employing frequency domain guidance gating edge enhancement, carrying out the global context modeling of an enhanced feature graph based on a multi-head self-attention mechanism, and carrying out the detection of a multi-dimensional image; after residual connection and layer normalization, the features are sent to a feed-forward network for nonlinear transformation, and high-dimensional features subjected to global coding are output; the method comprises the following steps: acquiring target query of information interaction based on self-attention, performing cross-attention-based interaction on the target query and high-dimensional features, then performing channel and space weighting processing on the interacted high-dimensional features through lightweight multi-dimensional attention, mapping the high-dimensional features into a target category and bounding box coordinates through a prediction head, and outputting a final detection result; the problems that a remote sensing image is complex in background, large in target scale difference and low in small target pixel proportion, and inherent speckle noise exists in a synthetic aperture radar image are solved.
Need to check novelty before this filing date? Find Prior Art