无线电工程2026，Vol.56Issue(2)：326-334,9.DOI:10.3969/j.issn.1003-3106.2026.02.015
GAENet:复杂背景下遥感图像的舰船旋转目标检测网络
GAENet:Network for Rotated Object Detection of Ships in Complex Remote Sensing Images
摘要
Abstract
Achieving high-precision detection of rotated ship targets under complex background conditions remains a critical technical challenge in maritime target recognition within the field of computer vision.Affected by factors such as multi-scale target variations,severe background interference,dense target distributions,and arbitrary orientations,there is an urgent need to design ship-oriented detection algorithms with strong robustness and high discriminative capability to improve detection performance and generalization in real-world scenarios.In this work,an Adaptive Channel Attention Mechanism(ACAM)is introduced,which enables the network to automatically focus on regions and targets that are more critical for classification and localization tasks,thereby effectively integrating global and local information of ship targets in remote sensing images.Meanwhile,a lightweight Grouped Efficient Feature Fusion(GEFF)module is designed and combined with a novel U-shaped Feature Pyramid Network(U-FPN),allowing the fusion process to fully exploit the global and local receptive field information provided by ACAM.Furthermore,a Shared Task Dynamic Alignment Detection Head(STADH)is proposed to enhance the collaborative optimization between classification and regression tasks.Experimental results on the SCOSS and HRSC2016 datasets demonstrate that,compared with conventional methods,the proposed approach improves the mean Average Precision(mAP)by 1.4%and 1.1%,respectively,while reducing the number of parameters by 40%and the computational cost by 15.6%.关键词
遥感图像/舰船目标检测/轻量化/多尺度特征融合/旋转目标Key words
remote sensing images/ship Rotated object detection/lightweight/multi-scale feature fusion/rotating target分类
信息技术与安全科学引用本文复制引用
张向朝,彭冬亮,罗昕,陈锘..GAENet:复杂背景下遥感图像的舰船旋转目标检测网络[J].无线电工程,2026,56(2):326-334,9.基金项目
浙江省自然科学基金重点项目(LZ23F030002) Zhejiang Provincial Natural Science Foundation of China(LZ23F030002) （LZ23F030002）
