Remote Sensing Small Object Detection Network Based on Attention Mechanism and Multi-Scale Feature Fusion
In remote sensing images, small objects have too few discriminative features, are easily confused with background information, and are difficult to locate, leading to a degradation in detection accuracy when using general object detection networks for aerial images. To solve the above problems, we p...
Main Authors: | , , , , |
---|---|
Format: | Article |
Language: | English |
Published: |
MDPI AG
2023-05-01
|
Series: | Remote Sensing |
Subjects: | |
Online Access: | https://www.mdpi.com/2072-4292/15/11/2728 |
_version_ | 1797596855444439040 |
---|---|
author | Junsuo Qu Zongbing Tang Le Zhang Yanghai Zhang Zhenguo Zhang |
author_facet | Junsuo Qu Zongbing Tang Le Zhang Yanghai Zhang Zhenguo Zhang |
author_sort | Junsuo Qu |
collection | DOAJ |
description | In remote sensing images, small objects have too few discriminative features, are easily confused with background information, and are difficult to locate, leading to a degradation in detection accuracy when using general object detection networks for aerial images. To solve the above problems, we propose a remote sensing small object detection network based on the attention mechanism and multi-scale feature fusion, and name it AMMFN. Firstly, a detection head enhancement module (DHEM) was designed to strengthen the characterization of small object features through a combination of multi-scale feature fusion and attention mechanisms. Secondly, an attention mechanism based channel cascade (AMCC) module was designed to reduce the redundant information in the feature layer and protect small objects from information loss during feature fusion. Then, the Normalized Wasserstein Distance (NWD) was introduced and combined with Generalized Intersection over Union (GIoU) as the location regression loss function to improve the optimization weight of the model for small objects and the accuracy of the regression boxes. Finally, an object detection layer was added to improve the object feature extraction ability at different scales. Experimental results from the Unmanned Aerial Vehicles (UAV) dataset VisDrone2021 and the homemade dataset show that the AMMFN improves the <inline-formula><math xmlns="http://www.w3.org/1998/Math/MathML" display="inline"><semantics><mrow><msub><mrow><mi mathvariant="normal">A</mi><mi mathvariant="normal">P</mi></mrow><mrow><mi mathvariant="normal">s</mi></mrow></msub></mrow></semantics></math></inline-formula> values by 2.4% and 3.2%, respectively, compared with YOLOv5s, which represents an effective improvement in the detection accuracy of small objects. |
first_indexed | 2024-03-11T02:58:54Z |
format | Article |
id | doaj.art-bc11e27d8a6a4569a0ee77390d7ca658 |
institution | Directory Open Access Journal |
issn | 2072-4292 |
language | English |
last_indexed | 2024-03-11T02:58:54Z |
publishDate | 2023-05-01 |
publisher | MDPI AG |
record_format | Article |
series | Remote Sensing |
spelling | doaj.art-bc11e27d8a6a4569a0ee77390d7ca6582023-11-18T08:27:52ZengMDPI AGRemote Sensing2072-42922023-05-011511272810.3390/rs15112728Remote Sensing Small Object Detection Network Based on Attention Mechanism and Multi-Scale Feature FusionJunsuo Qu0Zongbing Tang1Le Zhang2Yanghai Zhang3Zhenguo Zhang4Xi’an Key Laboratory of Advanced Control and Intelligent Process, School of Automation, Xi’an Robertic Intelligent Systems International Science and Technology Cooperation Base, Xi’an University of Posts and Telecommunications, Xi’an 710121, ChinaXi’an Key Laboratory of Advanced Control and Intelligent Process, School of Automation, Xi’an Robertic Intelligent Systems International Science and Technology Cooperation Base, Xi’an University of Posts and Telecommunications, Xi’an 710121, ChinaXi’an Key Laboratory of Advanced Control and Intelligent Process, School of Automation, Xi’an Robertic Intelligent Systems International Science and Technology Cooperation Base, Xi’an University of Posts and Telecommunications, Xi’an 710121, ChinaXi’an Key Laboratory of Advanced Control and Intelligent Process, School of Automation, Xi’an Robertic Intelligent Systems International Science and Technology Cooperation Base, Xi’an University of Posts and Telecommunications, Xi’an 710121, ChinaXi’an Key Laboratory of Advanced Control and Intelligent Process, School of Automation, Xi’an Robertic Intelligent Systems International Science and Technology Cooperation Base, Xi’an University of Posts and Telecommunications, Xi’an 710121, ChinaIn remote sensing images, small objects have too few discriminative features, are easily confused with background information, and are difficult to locate, leading to a degradation in detection accuracy when using general object detection networks for aerial images. To solve the above problems, we propose a remote sensing small object detection network based on the attention mechanism and multi-scale feature fusion, and name it AMMFN. Firstly, a detection head enhancement module (DHEM) was designed to strengthen the characterization of small object features through a combination of multi-scale feature fusion and attention mechanisms. Secondly, an attention mechanism based channel cascade (AMCC) module was designed to reduce the redundant information in the feature layer and protect small objects from information loss during feature fusion. Then, the Normalized Wasserstein Distance (NWD) was introduced and combined with Generalized Intersection over Union (GIoU) as the location regression loss function to improve the optimization weight of the model for small objects and the accuracy of the regression boxes. Finally, an object detection layer was added to improve the object feature extraction ability at different scales. Experimental results from the Unmanned Aerial Vehicles (UAV) dataset VisDrone2021 and the homemade dataset show that the AMMFN improves the <inline-formula><math xmlns="http://www.w3.org/1998/Math/MathML" display="inline"><semantics><mrow><msub><mrow><mi mathvariant="normal">A</mi><mi mathvariant="normal">P</mi></mrow><mrow><mi mathvariant="normal">s</mi></mrow></msub></mrow></semantics></math></inline-formula> values by 2.4% and 3.2%, respectively, compared with YOLOv5s, which represents an effective improvement in the detection accuracy of small objects.https://www.mdpi.com/2072-4292/15/11/2728small object detectionattention mechanismloss functionremote sensing |
spellingShingle | Junsuo Qu Zongbing Tang Le Zhang Yanghai Zhang Zhenguo Zhang Remote Sensing Small Object Detection Network Based on Attention Mechanism and Multi-Scale Feature Fusion Remote Sensing small object detection attention mechanism loss function remote sensing |
title | Remote Sensing Small Object Detection Network Based on Attention Mechanism and Multi-Scale Feature Fusion |
title_full | Remote Sensing Small Object Detection Network Based on Attention Mechanism and Multi-Scale Feature Fusion |
title_fullStr | Remote Sensing Small Object Detection Network Based on Attention Mechanism and Multi-Scale Feature Fusion |
title_full_unstemmed | Remote Sensing Small Object Detection Network Based on Attention Mechanism and Multi-Scale Feature Fusion |
title_short | Remote Sensing Small Object Detection Network Based on Attention Mechanism and Multi-Scale Feature Fusion |
title_sort | remote sensing small object detection network based on attention mechanism and multi scale feature fusion |
topic | small object detection attention mechanism loss function remote sensing |
url | https://www.mdpi.com/2072-4292/15/11/2728 |
work_keys_str_mv | AT junsuoqu remotesensingsmallobjectdetectionnetworkbasedonattentionmechanismandmultiscalefeaturefusion AT zongbingtang remotesensingsmallobjectdetectionnetworkbasedonattentionmechanismandmultiscalefeaturefusion AT lezhang remotesensingsmallobjectdetectionnetworkbasedonattentionmechanismandmultiscalefeaturefusion AT yanghaizhang remotesensingsmallobjectdetectionnetworkbasedonattentionmechanismandmultiscalefeaturefusion AT zhenguozhang remotesensingsmallobjectdetectionnetworkbasedonattentionmechanismandmultiscalefeaturefusion |