Small Object Detection Algorithm for Aerial Photography Based on Improved YOLOv3

This study presents an improved You Only Look Once version 3 (YOLOv3) algorithm for small object detection, to address problems such as low detection precision for small objects, missed detection, and false detection in the detection process. First, in terms of network structure, the feature extract...

Ausführliche Beschreibung

Gespeichert in:
Bibliographische Detailangaben
Veröffentlicht in:Ji suan ji gong cheng Jg. 51; H. 6; S. 184 - 192
1. Verfasser: XI Qi, WANG Mingjie, WEI Jinghe, ZHAO Wei
Format: Journal Article
Sprache:Chinesisch
Englisch
Veröffentlicht: Editorial Office of Computer Engineering 15.06.2025
Schlagworte:
small object detection|you only look once version 3 (yolov3)|densely connected network (densenet)|loss function|generalized intersection over union (giou)
ISSN:1000-3428
Online-Zugang:Volltext
Tags: Tag hinzufügen
Keine Tags, Fügen Sie den ersten Tag hinzu!
Abstract This study presents an improved You Only Look Once version 3 (YOLOv3) algorithm for small object detection, to address problems such as low detection precision for small objects, missed detection, and false detection in the detection process. First, in terms of network structure, the feature extraction capability of the backbone network is improved by using DenseNet-121, with a Densely Connected Network (DenseNet), to replace the original Darknet-53 network as its basic network. Simultaneously, the convolution kernel size is modified to further reduce the loss of feature map information, to enhance the robustness of the detection model against small objects. A fourth feature detection layer with a size of 104×104 pixel is added. Second, the bilinear interpolation method is used to replace the original nearest neighbor interpolation method for upsampling operations, to solve the serious feature loss problem in most detection algorithms. Finally, in terms of the loss function, Generalized Intersection over Union (GIoU) is used instead of Intersection over Union (IoU) to calculate the loss value of the boundary frame, and the Focal Loss function is introduced as the confidence loss function of the boundary frame. Experimental results show that the mean Average Precision (mAP) of the improved algorithm on the VisDrone2019 dataset is 63.3%, which is 13.2 percentage points higher than that of the original YOLOv3 detection model, and 52 frame/s on a GTX 1080 Ti device. The improved algorithm has good detection performance for small objects.
AbstractList This study presents an improved You Only Look Once version 3 (YOLOv3) algorithm for small object detection, to address problems such as low detection precision for small objects, missed detection, and false detection in the detection process. First, in terms of network structure, the feature extraction capability of the backbone network is improved by using DenseNet-121, with a Densely Connected Network (DenseNet), to replace the original Darknet-53 network as its basic network. Simultaneously, the convolution kernel size is modified to further reduce the loss of feature map information, to enhance the robustness of the detection model against small objects. A fourth feature detection layer with a size of 104×104 pixel is added. Second, the bilinear interpolation method is used to replace the original nearest neighbor interpolation method for upsampling operations, to solve the serious feature loss problem in most detection algorithms. Finally, in terms of the loss function, Generalized Intersection over Union (GIoU) is used instead of Intersection over Union (IoU) to calculate the loss value of the boundary frame, and the Focal Loss function is introduced as the confidence loss function of the boundary frame. Experimental results show that the mean Average Precision (mAP) of the improved algorithm on the VisDrone2019 dataset is 63.3%, which is 13.2 percentage points higher than that of the original YOLOv3 detection model, and 52 frame/s on a GTX 1080 Ti device. The improved algorithm has good detection performance for small objects.
Author XI Qi, WANG Mingjie, WEI Jinghe, ZHAO Wei
Author_xml – sequence: 1
  fullname: XI Qi, WANG Mingjie, WEI Jinghe, ZHAO Wei
  organization: The 58th Research Institute of China Electronics Technology Group Corporation, Wuxi 214122, Jiangsu, China
BookMark eNo9jM1OwzAQhH0oEi3wDuYBEjZex3aOpfxVqpRKwIFTtImdNlESV05UqW9PBIjTzHyamRVbDH5wjN0nECeZ0uahjZtxHOIEACKUwsQAyqjMLNjyn12z1Ti2AFIIgCXbv_fUdTwvW1dN_MlNszR-4Ovu4EMzHXte-8DXLjTU8f3RT_4Q6HS88EcaneVzc9ufgj_P_ivf5We8ZVc1daO7-9Mb9vny_LF5i3b563az3kU2QTNFCp1VlIBVQptUaHIOgYQhLAUpVGmVWKxBWlIZSlfWwqLTJNFqNccEb9j299d6aotTaHoKl8JTU_wAHw4FhampOlcAVlChS41JSWbz2qlSu0rUoGtZaonfb7tgHg
ContentType Journal Article
DBID DOA
DOI 10.19678/j.issn.1000-3428.0068698
DatabaseName DOAJ Directory of Open Access Journals
DatabaseTitleList
Database_xml – sequence: 1
  dbid: DOA
  name: DOAJ Directory of Open Access Journals
  url: https://www.doaj.org/
  sourceTypes: Open Website
DeliveryMethod fulltext_linktorsrc
Discipline Engineering
EndPage 192
ExternalDocumentID oai_doaj_org_article_03c0c3e5885a494ebe6b7ec2f07f4b74
GroupedDBID -0Y
92H
92I
ABJNI
ACGFS
ADMLS
ALMA_UNASSIGNED_HOLDINGS
CCEZO
CUBFJ
CW9
GROUPED_DOAJ
TCJ
TGT
U1G
U5S
ID FETCH-LOGICAL-d138t-63ed6a10d6278527aee30a28a3b2a6365c1d3f04da6934ebf2d3e7a43d764eb13
IEDL.DBID DOA
ISSN 1000-3428
IngestDate Mon Nov 03 22:01:13 EST 2025
IsDoiOpenAccess true
IsOpenAccess true
IsPeerReviewed true
IsScholarly true
Issue 6
Language Chinese
English
LinkModel DirectLink
MergedId FETCHMERGED-LOGICAL-d138t-63ed6a10d6278527aee30a28a3b2a6365c1d3f04da6934ebf2d3e7a43d764eb13
OpenAccessLink https://doaj.org/article/03c0c3e5885a494ebe6b7ec2f07f4b74
PageCount 9
ParticipantIDs doaj_primary_oai_doaj_org_article_03c0c3e5885a494ebe6b7ec2f07f4b74
PublicationCentury 2000
PublicationDate 2025-06-15
PublicationDateYYYYMMDD 2025-06-15
PublicationDate_xml – month: 06
  year: 2025
  text: 2025-06-15
  day: 15
PublicationDecade 2020
PublicationTitle Ji suan ji gong cheng
PublicationYear 2025
Publisher Editorial Office of Computer Engineering
Publisher_xml – name: Editorial Office of Computer Engineering
SSID ssj0042200
Score 2.2948334
Snippet This study presents an improved You Only Look Once version 3 (YOLOv3) algorithm for small object detection, to address problems such as low detection precision...
SourceID doaj
SourceType Open Website
StartPage 184
SubjectTerms small object detection|you only look once version 3 (yolov3)|densely connected network (densenet)|loss function|generalized intersection over union (giou)
Title Small Object Detection Algorithm for Aerial Photography Based on Improved YOLOv3
URI https://doaj.org/article/03c0c3e5885a494ebe6b7ec2f07f4b74
Volume 51
hasFullText 1
inHoldings 1
isFullTextHit
isPrint
journalDatabaseRights – providerCode: PRVAON
  databaseName: DOAJ Directory of Open Access Journals
  issn: 1000-3428
  databaseCode: DOA
  dateStart: 20160101
  customDbUrl:
  isFulltext: true
  dateEnd: 99991231
  titleUrlDefault: https://www.doaj.org/
  omitProxy: false
  ssIdentifier: ssj0042200
  providerName: Directory of Open Access Journals
link http://cvtisr.summon.serialssolutions.com/2.0.0/link/0/eLvHCXMwrV3LS8MwHA4yRPQgPvFNBK_Z0rx73NThQbaBCvNU0iR1ytbJrPv7TdIq8-TFW9OGNvy-5vcIyfcBcGWNLogVCZIJKRCTskA5dRgxF-KJdYoLE8Um5GCgxuN0tCL1FfaE1fTAteE6mBpsqONKcc1S5r8pcukMKbAsWC4jEyiW6XcxVftgRgiueQiw9zI-w94Al_FknnfNnbc4vdo_z9rxmESqfrH2x_DS3wHbTV4Iu_V4dsGaK_fA1gpb4D4YPcz0dAqHeVg6gTeuiruoStidvsx9iT-ZQZ-Awm78peBoMq8aNmrY85HKQt-zXkHw18_D--GSHoCn_u3j9R1qBBGQTaiqkKDOCp1gK4hUnEjtHMWaKE1zogUV3CSWFphZLVLqTeVRoE5qRq0UvpnQQ9Aq56U7AlDaVBpfyujcUFYk3OeN3FhLiLU4dD0GvWCM7L3mvMgCC3W84bHJGmyyv7A5-Y-XnIJNEjR3g14QPwOtavHpzsG6WVavH4uLCPsXpG6wyw
linkProvider Directory of Open Access Journals
openUrl ctx_ver=Z39.88-2004&ctx_enc=info%3Aofi%2Fenc%3AUTF-8&rfr_id=info%3Asid%2Fsummon.serialssolutions.com&rft_val_fmt=info%3Aofi%2Ffmt%3Akev%3Amtx%3Ajournal&rft.genre=article&rft.atitle=Small+Object+Detection+Algorithm+for+Aerial+Photography+Based+on+Improved+YOLOv3&rft.jtitle=Ji+suan+ji+gong+cheng&rft.au=XI+Qi%2C+WANG+Mingjie%2C+WEI+Jinghe%2C+ZHAO+Wei&rft.date=2025-06-15&rft.pub=Editorial+Office+of+Computer+Engineering&rft.issn=1000-3428&rft.volume=51&rft.issue=6&rft.spage=184&rft.epage=192&rft_id=info:doi/10.19678%2Fj.issn.1000-3428.0068698&rft.externalDBID=DOA&rft.externalDocID=oai_doaj_org_article_03c0c3e5885a494ebe6b7ec2f07f4b74
thumbnail_l http://covers-cdn.summon.serialssolutions.com/index.aspx?isbn=/lc.gif&issn=1000-3428&client=summon
thumbnail_m http://covers-cdn.summon.serialssolutions.com/index.aspx?isbn=/mc.gif&issn=1000-3428&client=summon
thumbnail_s http://covers-cdn.summon.serialssolutions.com/index.aspx?isbn=/sc.gif&issn=1000-3428&client=summon