Remote Sensing Small Object Detection Based on Multicontextual Information Aggregation

Due to wide field of view and background confusion, remote sensing objects are small and densely packed, hence commonly used detection methods detecting small objects are not satisfactory. In this article, we propose the multicontextual information aggregation YOLO (MCIA-YOLO) method, combining thre...

Full description

Saved in:
Bibliographic Details
Main Authors: Jingyu Wang, Mingrui Ma, Pengfei Huang, Shaohui Mei, Liang Zhang, Hongmei Wang
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/10891715/
Tags: Add Tag
No Tags, Be the first to tag this record!
_version_ 1850276198887718912
author Jingyu Wang
Mingrui Ma
Pengfei Huang
Shaohui Mei
Liang Zhang
Hongmei Wang
author_facet Jingyu Wang
Mingrui Ma
Pengfei Huang
Shaohui Mei
Liang Zhang
Hongmei Wang
author_sort Jingyu Wang
collection DOAJ
description Due to wide field of view and background confusion, remote sensing objects are small and densely packed, hence commonly used detection methods detecting small objects are not satisfactory. In this article, we propose the multicontextual information aggregation YOLO (MCIA-YOLO) method, combining three novel modules to effectively aggregate multicontextual information across channels, depths, and pixels. First, the channel-spatial information aggregation module assembles spatial global features pursuant to channel contextual information, increasing the density of key information. Second, the shallow-deep information sparse aggregation module applies a sparse cross self-attention mechanism. By sparsely correlating long-range dependency information across different regions, the representation capability of a small target is enhanced while removing redundant information. Third, to enrich local multiscale features and better identify dense targets, multiscale weighted aggregation module convolves multireceptive field information and performs weighted fusion. Our method demonstrates satisfactory performance on dataset VisDrone2019, UAVDT, and NWPU VHR-10, especially in small objects detection, surpassing several state-of-the-art methods.
format Article
id doaj-art-e2776f65718a490d86e125c1ed5ee2ce
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-e2776f65718a490d86e125c1ed5ee2ce2025-08-20T01:50:22ZengIEEEIEEE Journal of Selected Topics in Applied Earth Observations and Remote Sensing1939-14042151-15352025-01-01188248826010.1109/JSTARS.2025.354318910891715Remote Sensing Small Object Detection Based on Multicontextual Information AggregationJingyu Wang0https://orcid.org/0000-0001-7017-1938Mingrui Ma1https://orcid.org/0009-0000-4078-2391Pengfei Huang2https://orcid.org/0000-0003-4092-2918Shaohui Mei3https://orcid.org/0000-0002-8018-596XLiang Zhang4Hongmei Wang5https://orcid.org/0000-0002-1108-3505School of Astronautics, Northwestern Polytechnical University, Xi'an, ChinaUnmanned System Research Institute, Northwestern Polytechnical University, Xi'an, ChinaSchool of Astronautics, Northwestern Polytechnical University, Xi'an, ChinaSchool of Electronics and Information, Northwestern Polytechnical University, Xi'an, ChinaNational Key Laboratory of Air-Based Information Perception and Fusion, Luoyang, ChinaSchool of Astronautics, Northwestern Polytechnical University, Xi'an, ChinaDue to wide field of view and background confusion, remote sensing objects are small and densely packed, hence commonly used detection methods detecting small objects are not satisfactory. In this article, we propose the multicontextual information aggregation YOLO (MCIA-YOLO) method, combining three novel modules to effectively aggregate multicontextual information across channels, depths, and pixels. First, the channel-spatial information aggregation module assembles spatial global features pursuant to channel contextual information, increasing the density of key information. Second, the shallow-deep information sparse aggregation module applies a sparse cross self-attention mechanism. By sparsely correlating long-range dependency information across different regions, the representation capability of a small target is enhanced while removing redundant information. Third, to enrich local multiscale features and better identify dense targets, multiscale weighted aggregation module convolves multireceptive field information and performs weighted fusion. Our method demonstrates satisfactory performance on dataset VisDrone2019, UAVDT, and NWPU VHR-10, especially in small objects detection, surpassing several state-of-the-art methods.https://ieeexplore.ieee.org/document/10891715/Global contextual informationmulticontextual informationmultireceptive field enhancementsmall object detectionsparse cross self-attention
spellingShingle Jingyu Wang
Mingrui Ma
Pengfei Huang
Shaohui Mei
Liang Zhang
Hongmei Wang
Remote Sensing Small Object Detection Based on Multicontextual Information Aggregation
IEEE Journal of Selected Topics in Applied Earth Observations and Remote Sensing
Global contextual information
multicontextual information
multireceptive field enhancement
small object detection
sparse cross self-attention
title Remote Sensing Small Object Detection Based on Multicontextual Information Aggregation
title_full Remote Sensing Small Object Detection Based on Multicontextual Information Aggregation
title_fullStr Remote Sensing Small Object Detection Based on Multicontextual Information Aggregation
title_full_unstemmed Remote Sensing Small Object Detection Based on Multicontextual Information Aggregation
title_short Remote Sensing Small Object Detection Based on Multicontextual Information Aggregation
title_sort remote sensing small object detection based on multicontextual information aggregation
topic Global contextual information
multicontextual information
multireceptive field enhancement
small object detection
sparse cross self-attention
url https://ieeexplore.ieee.org/document/10891715/
work_keys_str_mv AT jingyuwang remotesensingsmallobjectdetectionbasedonmulticontextualinformationaggregation
AT mingruima remotesensingsmallobjectdetectionbasedonmulticontextualinformationaggregation
AT pengfeihuang remotesensingsmallobjectdetectionbasedonmulticontextualinformationaggregation
AT shaohuimei remotesensingsmallobjectdetectionbasedonmulticontextualinformationaggregation
AT liangzhang remotesensingsmallobjectdetectionbasedonmulticontextualinformationaggregation
AT hongmeiwang remotesensingsmallobjectdetectionbasedonmulticontextualinformationaggregation