A Semantic Feature Enhancement-Based Aerial Image Target Detection Method Using Dense RFB-FE

Author:

Li Xinyang1,Zhang Jingguo1

Affiliation:

1. Changchun Institute of Optics, Fine Mechanics and Physics, Chinese Academy of Sciences, Changchun, China

Abstract

Aerial image target detection is a challenging task due to the complex backgrounds, dense target distribution, and large-scale differences often present in aerial images. Existing methods often struggle to effectively extract detailed features and address the issue of imbalanced positive and negative samples. To tackle these challenges, an aerial image target detection method (dense RFB-FE-CGAM) based on dense RFB-FE and channel-global attention mechanism (CGAM) was proposed. First, the authors design a shallow feature enhancement module using dense RFB feature multiplexing and expand convolution within an SSD network, improving detailed feature extraction. Second, they introduce CGAM, a global attention module, to enhance semantic feature extraction in backbone networks. Finally, they incorporate a focal loss function for joint training, addressing sample imbalance. In experiments, the method achieved an mAP of 0.755 on the DOTA dataset and recall/AP values of 0.889/0.906 on HRSC2016, confirming the effectiveness of dense RFB-FE-CGAM for aerial image target detection.

Publisher

IGI Global

Subject

Computer Networks and Communications,Information Systems

Cited by 1 articles. 订阅此论文施引文献 订阅此论文施引文献,注册后可以免费订阅5篇论文的施引文献,订阅后可以查看论文全部施引文献

1. YOLO-DCNet;International Journal on Semantic Web and Information Systems;2024-02-21

同舟云学术

1.学者识别学者识别

2.学术分析学术分析

3.人才评估人才评估

"同舟云学术"是以全球学者为主线,采集、加工和组织学术论文而形成的新型学术文献查询和分析系统,可以对全球学者进行文献检索和人才价值评估。用户可以通过关注某些学科领域的顶尖人物而持续追踪该领域的学科进展和研究前沿。经过近期的数据扩容,当前同舟云学术共收录了国内外主流学术期刊6万余种,收集的期刊论文及会议论文总量共计约1.5亿篇,并以每天添加12000余篇中外论文的速度递增。我们也可以为用户提供个性化、定制化的学者数据。欢迎来电咨询!咨询电话:010-8811{复制后删除}0370

www.globalauthorid.com

TOP

Copyright © 2019-2024 北京同舟云网络信息技术有限公司
京公网安备11010802033243号  京ICP备18003416号-3