JOURNAL ARTICLE

Optical remote sensing image object detection based on multi-resolution feature fusion

Yanqing YaoGong ChengXingxing XieJunwei Han

Year: 2021 Journal:   National Remote Sensing Bulletin Vol: 25 (5)Pages: 1124-1137   Publisher: Science Press

Abstract

高分辨率遥感图像目标检测是计算机视觉的一个重要研究领域,在民用与军事领域具有重要的应用价值。目前,基于深度学习的自然图像目标检测有了突破性进展。但是,由于遥感图像具有目标尺度差异大且类间相似度高的特点,使得处理自然图像的目标检测算法直接应用于遥感图像时仍面临着一些挑战。针对上述挑战,本文提出一种多分辨率特征融合的遥感图像目标检测方法。首先,通过特征金字塔提取多尺度特征图并在其后嵌入多分辨率特征提取网络,促使网络学习目标在不同分辨率下的特征,缩小不同特征层之间的语义差距。其次,为实现多分辨特征的有效融合,本文采用自适应特征融合模块挖掘更具判别性的多分辨特征表达。最后,将自适应特征融合模块的输出特征的相邻层进行深度融合。在公开的遥感图像目标检测数据集DIOR和DOTA上评估了本文方法的有效性,相比采用特征金字塔结构的Faster R-CNN,本文方法的准确率(mAP)分别提高2.5%和2.2%。

Keywords:
Computer vision Artificial intelligence Remote sensing Image fusion Feature (linguistics) Computer science Fusion Object (grammar) Object detection Object based Image (mathematics) Geology Pattern recognition (psychology)

Metrics

24
Cited By
2.06
FWCI (Field Weighted Citation Impact)
35
Refs
0.89
Citation Normalized Percentile
Is in top 1%
Is in top 10%

Citation History

Topics

Remote Sensing and Land Use
Physical Sciences →  Earth and Planetary Sciences →  Atmospheric Science
Remote-Sensing Image Classification
Physical Sciences →  Engineering →  Media Technology
Environmental Changes in China
Physical Sciences →  Environmental Science →  Global and Planetary Change
© 2026 ScienceGate Book Chapters — All rights reserved.