SODRS: Semisupervised Learning for One-Stage Small Object Detection in Remote Sensing Images

Small object detection in remote sensing images faces challenges such as weak features, vulnerability to interference, and limited object visibility. These factors have made it a long-standing technical issue. Traditional object detection methods often perform poorly in this field, especially when d...

Full description

Saved in:
Bibliographic Details
Main Authors: Mingquan Liu, Lei Kuang, Chengjun Li, Jing Tian, Zifang Chen, Xuewu Han
Format: Article
Language:English
Published: IEEE 2025-01-01
Series:IEEE Journal of Selected Topics in Applied Earth Observations and Remote Sensing
Subjects:
Online Access:https://ieeexplore.ieee.org/document/10947527/
Tags: Add Tag
No Tags, Be the first to tag this record!
_version_ 1850173711213133824
author Mingquan Liu
Lei Kuang
Chengjun Li
Jing Tian
Zifang Chen
Xuewu Han
author_facet Mingquan Liu
Lei Kuang
Chengjun Li
Jing Tian
Zifang Chen
Xuewu Han
author_sort Mingquan Liu
collection DOAJ
description Small object detection in remote sensing images faces challenges such as weak features, vulnerability to interference, and limited object visibility. These factors have made it a long-standing technical issue. Traditional object detection methods often perform poorly in this field, especially when dealing with high-resolution remote sensing images, where detection accuracy and robustness fail to meet practical application needs. Furthermore, due to the large volume of remote sensing image data and the high cost of annotation, effectively utilizing unlabeled data has become the key to enhancing model detection performance. To resolve these problems, this article proposes a one-stage semisupervised learning method for small object detection in remote sensing images called SODRS. This method uses FCOS as the baseline and introduces the squeeze-and-excitation attention mechanism to enhance feature representation, the soft focal loss module to optimize the class ambiguity between objects and background, and the confident pseudolabeling strategy to improve the quality of pseudolabels. Finally, the conditional random fields-based label refinement is applied to postprocess the predicted labels, improving spatial relationships and dependencies among objects. Experimental results demonstrate that SODRS excels at detecting small objects in complex remote sensing scenarios, with higher accuracy and robustness than existing classical methods. Notably, SODRS can accurately distinguish objects even when densely distributed, close to each other, or overlapping. This demonstrates the application potential of one-stage semisupervised methods in small object detection in remote sensing images.
format Article
id doaj-art-f558c8d4e44e4c72b0ed11ce6f77b0cb
institution OA Journals
issn 1939-1404
2151-1535
language English
publishDate 2025-01-01
publisher IEEE
record_format Article
series IEEE Journal of Selected Topics in Applied Earth Observations and Remote Sensing
spelling doaj-art-f558c8d4e44e4c72b0ed11ce6f77b0cb2025-08-20T02:19:47ZengIEEEIEEE Journal of Selected Topics in Applied Earth Observations and Remote Sensing1939-14042151-15352025-01-0118107111072310.1109/JSTARS.2025.355709210947527SODRS: Semisupervised Learning for One-Stage Small Object Detection in Remote Sensing ImagesMingquan Liu0https://orcid.org/0009-0009-2636-4869Lei Kuang1https://orcid.org/0009-0000-9440-736XChengjun Li2https://orcid.org/0000-0002-0779-622XJing Tian3Zifang Chen4Xuewu Han5School of Computer Science, China University of Geosciences, Wuhan, ChinaSchool of Computer Science, China University of Geosciences, Wuhan, ChinaSchool of Computer Science, China University of Geosciences, Wuhan, ChinaChangjiang Institute of Survey, Planning, Design and Research, Wuhan, ChinaSchool of Computer Science, China University of Geosciences, Wuhan, ChinaState Grid Electric Power Research Institute Wuhan NARI Company Ltd., Wuhan, ChinaSmall object detection in remote sensing images faces challenges such as weak features, vulnerability to interference, and limited object visibility. These factors have made it a long-standing technical issue. Traditional object detection methods often perform poorly in this field, especially when dealing with high-resolution remote sensing images, where detection accuracy and robustness fail to meet practical application needs. Furthermore, due to the large volume of remote sensing image data and the high cost of annotation, effectively utilizing unlabeled data has become the key to enhancing model detection performance. To resolve these problems, this article proposes a one-stage semisupervised learning method for small object detection in remote sensing images called SODRS. This method uses FCOS as the baseline and introduces the squeeze-and-excitation attention mechanism to enhance feature representation, the soft focal loss module to optimize the class ambiguity between objects and background, and the confident pseudolabeling strategy to improve the quality of pseudolabels. Finally, the conditional random fields-based label refinement is applied to postprocess the predicted labels, improving spatial relationships and dependencies among objects. Experimental results demonstrate that SODRS excels at detecting small objects in complex remote sensing scenarios, with higher accuracy and robustness than existing classical methods. Notably, SODRS can accurately distinguish objects even when densely distributed, close to each other, or overlapping. This demonstrates the application potential of one-stage semisupervised methods in small object detection in remote sensing images.https://ieeexplore.ieee.org/document/10947527/Feature enhancementlabel refinementone-stage semisupervised learning (SSL)remote sensing imagessmall object detection
spellingShingle Mingquan Liu
Lei Kuang
Chengjun Li
Jing Tian
Zifang Chen
Xuewu Han
SODRS: Semisupervised Learning for One-Stage Small Object Detection in Remote Sensing Images
IEEE Journal of Selected Topics in Applied Earth Observations and Remote Sensing
Feature enhancement
label refinement
one-stage semisupervised learning (SSL)
remote sensing images
small object detection
title SODRS: Semisupervised Learning for One-Stage Small Object Detection in Remote Sensing Images
title_full SODRS: Semisupervised Learning for One-Stage Small Object Detection in Remote Sensing Images
title_fullStr SODRS: Semisupervised Learning for One-Stage Small Object Detection in Remote Sensing Images
title_full_unstemmed SODRS: Semisupervised Learning for One-Stage Small Object Detection in Remote Sensing Images
title_short SODRS: Semisupervised Learning for One-Stage Small Object Detection in Remote Sensing Images
title_sort sodrs semisupervised learning for one stage small object detection in remote sensing images
topic Feature enhancement
label refinement
one-stage semisupervised learning (SSL)
remote sensing images
small object detection
url https://ieeexplore.ieee.org/document/10947527/
work_keys_str_mv AT mingquanliu sodrssemisupervisedlearningforonestagesmallobjectdetectioninremotesensingimages
AT leikuang sodrssemisupervisedlearningforonestagesmallobjectdetectioninremotesensingimages
AT chengjunli sodrssemisupervisedlearningforonestagesmallobjectdetectioninremotesensingimages
AT jingtian sodrssemisupervisedlearningforonestagesmallobjectdetectioninremotesensingimages
AT zifangchen sodrssemisupervisedlearningforonestagesmallobjectdetectioninremotesensingimages
AT xuewuhan sodrssemisupervisedlearningforonestagesmallobjectdetectioninremotesensingimages