FDBGF-Net: Frequency-Decoupled Bidirectional Gated Fusion for Infrared–Visible Object Detection Under Degradation in UAV Imagery

计算机科学 计算机视觉 目标检测 人工智能 降级(电信) 融合 对象(语法) 传感器融合 图像融合 特征提取 模式识别(心理学) 遥感 雷达探测 图像分割
作者
Kaipeng Wang,Guanglin He,Zongze Li,Xiaoyan Zheng,Hao Zhang
出处
期刊:IEEE Journal of Selected Topics in Applied Earth Observations and Remote Sensing [Institute of Electrical and Electronics Engineers]
卷期号:19: 36592-36611
标识
DOI:10.1109/jstars.2026.3746606
摘要

Infrared–visible fusion is increasingly used to improve object detection robustness in unmanned aerial vehicle (UAV) remote sensing. However, real-world UAV imagery is often affected by co-occurring degradations, such as blur, rain, snow, fog, strong illumination, and electromagnetic interference. These disturbances may simultaneously corrupt both modalities, weaken cross-modal complementarity, and make conventional fusion schemes unreliable. To address this problem, we propose FDBGF-Net, a frequency-decoupled bidirectional gated fusion network for infrared–visible object detection under compound degradation. The proposed framework introduces three coordinated components. First, a Dual-Modal Gated Fusion Network performs P3-level cross-modal alignment through selective state space modeling, preserving geometric details before high-level semantic compression. Second, a Degradation-Aware Gated Fusion Network employs a bidirectional gated feature pyramid and a spatial adaptive router to dynamically aggregate multi scale features according to modality reliability. Third, a Wavelet Decoupled Frequency Attention module separates fused representations into low-frequency structural and high-frequency detail components, and adaptively recalibrates their contributions using dual-path channel gating. We also construct a benchmark containing 3,711 infrared–visible image pairs, five target categories, six degradation types, and three severity levels. Experimental resolution and weaker texture information. Consequently, the complementary characteristics of the two modalities offer clear advantages for robust target detection. In practical UAV applications, however, infrared and visible images may be simultaneously affected by compound degradations, including blur, rain, snow, fog, strong illumination, and electromagnetic interference. Such degradations distort structural details and semantic responses in both modalities, weaken their complementarity, and make stable multimodal detection particularly challenging [3]–[6]. results show that FDBGF-Net achieves 79.8% mAP50 on the proposed benchmark, outperforming DPAL-R by 1.5 percentage points at 76.3 FPS. It also reaches 79.6% mAP50 on VEDAI and 88.1% mAP50 on Aligned FLIR after independent target-dataset retraining, demonstrating consistent cross-dataset applicability across substantially different data distributions.

科研通智能强力驱动
Strongly Powered by AbleSci AI
科研通是完全免费的文献互助平台,具备全网最快的应助速度,最高的求助完成率。 对每一个文献求助,科研通都将尽心尽力,给求助人一个满意的交代。
实时播报
杜胤江完成签到,获得积分10
1秒前
1秒前
su2589367856完成签到,获得积分10
1秒前
orixero的应助被feiyue126采纳,获得10
1秒前
REBECCA完成签到,获得积分10
1秒前
1秒前
J_Xu完成签到 ,获得积分10
1秒前
yjj发布了新的文献求助20
2秒前
lily发布了新的文献求助10
2秒前
ZZY完成签到,获得积分10
3秒前
爆米花的应助被科研通管家采纳,获得10
3秒前
陈哥完成签到,获得积分10
3秒前
Lucas的应助被科研通管家采纳,获得10
3秒前
顺顺利利的应助被科研通管家采纳,获得10
3秒前
Cyoka完成签到,获得积分10
3秒前
青橙的应助被科研通管家采纳,获得10
3秒前
罗lsz完成签到,获得积分10
3秒前
田様的应助被科研通管家采纳,获得10
3秒前
慢慢完成签到 ,获得积分10
3秒前
NexusExplorer的应助被科研通管家采纳,获得10
3秒前
少女闰土发布了新的文献求助10
4秒前
共享精神的应助被科研通管家采纳,获得10
4秒前
4秒前
可爱滴小花花完成签到,获得积分10
4秒前
秋风的应助被科研通管家采纳,获得10
4秒前
molihuakai的应助被科研通管家采纳,获得10
4秒前
飞飞的应助被科研通管家采纳,获得10
4秒前
方杰完成签到,获得积分10
4秒前
华仔的应助被科研通管家采纳,获得10
4秒前
4秒前
飞飞的应助被科研通管家采纳,获得10
4秒前
秋澄完成签到 ,获得积分10
5秒前
阿钦完成签到,获得积分10
5秒前
skmksd发布了新的文献求助10
5秒前
田様的应助被369ninja采纳,获得10
5秒前
5秒前
Chen完成签到,获得积分10
5秒前
HXX19完成签到,获得积分10
6秒前
Anothermoon完成签到,获得积分10
6秒前
sx完成签到,获得积分10
6秒前
高分求助中
(应助此贴封号)【重要!!请各用户(尤其是新用户)详细阅读】【科研通的精品贴汇总】 10000
拉伸高分子熔体中缠结行为的模拟和小角中子散射研究 1700
The World Atlas of Wine 9th edition 1000
奈特基础组织学:配套组织病理学(第4版)(Netter's Essential Histology With Correlated Histopathology) 800
John McDowells Bildungstheorie 800
Social Psychology in the Real World 600
八旗制度の研究 600
热门求助领域 (近24小时)
化学 材料科学 医学 生物 纳米技术 计算机科学 工程类 有机化学 化学工程 生物化学 物理 复合材料 内科学 催化作用 光电子学 人工智能 细胞生物学 无机化学 物理化学 数学
热门帖子
关注 科研通微信公众号,转发送积分 8180146
求助须知:如何正确求助?哪些是违规求助? 9570319
关于积分的说明 21809739
捐赠科研通 7776294
什么是DOI,文献DOI怎么找? 3433266
关于科研通互助平台的介绍 2557483
邀请新用户注册赠送积分活动 2479039