Improved detection algorithm of tank and armored vehicles based on YOLOV3-tiny

Zihan Zhao, Liu Peng
{"title":"Improved detection algorithm of tank and armored vehicles based on YOLOV3-tiny","authors":"Zihan Zhao, Liu Peng","doi":"10.1117/12.2639173","DOIUrl":null,"url":null,"abstract":"Target detection technology is a typical application in the military field. It can quickly and accurately find and identify all kinds of enemy vehicle targets in the battlefield, and respond to all kinds of battlefield targets more quickly, which has become the key to improve the battlefield situation. Because the battlefield environment is very complex, the traditional target detection algorithm is not ideal when detecting targets in complex scenes. Therefore, a target detection algorithm for armored vehicles of military tanks based on improved YOLOv3-tiny is proposed, which realizes the automatic detection of military targets in complex environments by deep learning. Firstly, based on YOLOv3-tiny algorithm, ResNext residual network is added to replace the original feature extraction network, which better improves the problem of missing and false detection of small targets and optimizes the convolution network structure. Then, the dense network is introduced, and the features of different layers are fused to realize feature reuse, which improves the efficiency of extracting better features of target vehicles, strengthens the network's ability to learn features, and improves the detection effect. Experimental results show that the recall rate and precision rate are increased by 4.62% and 3.79% respectively, the average precision rate is increased by 4.32%, and the frame rate can reach 78 frames/s.","PeriodicalId":336892,"journal":{"name":"Neural Networks, Information and Communication Engineering","volume":"5 1","pages":"0"},"PeriodicalIF":0.0000,"publicationDate":"2022-06-30","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":"0","resultStr":null,"platform":"Semanticscholar","paperid":null,"PeriodicalName":"Neural Networks, Information and Communication Engineering","FirstCategoryId":"1085","ListUrlMain":"https://doi.org/10.1117/12.2639173","RegionNum":0,"RegionCategory":null,"ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":null,"EPubDate":"","PubModel":"","JCR":"","JCRName":"","Score":null,"Total":0}
引用次数: 0

Abstract

Target detection technology is a typical application in the military field. It can quickly and accurately find and identify all kinds of enemy vehicle targets in the battlefield, and respond to all kinds of battlefield targets more quickly, which has become the key to improve the battlefield situation. Because the battlefield environment is very complex, the traditional target detection algorithm is not ideal when detecting targets in complex scenes. Therefore, a target detection algorithm for armored vehicles of military tanks based on improved YOLOv3-tiny is proposed, which realizes the automatic detection of military targets in complex environments by deep learning. Firstly, based on YOLOv3-tiny algorithm, ResNext residual network is added to replace the original feature extraction network, which better improves the problem of missing and false detection of small targets and optimizes the convolution network structure. Then, the dense network is introduced, and the features of different layers are fused to realize feature reuse, which improves the efficiency of extracting better features of target vehicles, strengthens the network's ability to learn features, and improves the detection effect. Experimental results show that the recall rate and precision rate are increased by 4.62% and 3.79% respectively, the average precision rate is increased by 4.32%, and the frame rate can reach 78 frames/s.
基于YOLOV3-tiny的坦克装甲车辆改进检测算法
目标探测技术是军事领域的典型应用。它能快速准确地发现和识别战场上的各种敌方车辆目标,对各种战场目标作出更快速的反应,成为改善战场态势的关键。由于战场环境非常复杂,传统的目标检测算法在检测复杂场景中的目标时并不理想。为此,提出了一种基于改进型YOLOv3-tiny的军用坦克装甲车目标检测算法,通过深度学习实现复杂环境下军事目标的自动检测。首先,在YOLOv3-tiny算法的基础上,加入ResNext残差网络代替原有的特征提取网络,更好地改善了小目标的缺失和误检问题,优化了卷积网络结构。然后,引入密集网络,融合不同层的特征,实现特征重用,提高了提取目标车辆更好特征的效率,增强了网络学习特征的能力,提高了检测效果。实验结果表明,该方法的查全率和查准率分别提高了4.62%和3.79%,平均查准率提高了4.32%,帧率达到78帧/秒。
本文章由计算机程序翻译,如有差异,请以英文原文为准。
求助全文
约1分钟内获得全文 求助全文
来源期刊
自引率
0.00%
发文量
0
×
引用
GB/T 7714-2015
复制
MLA
复制
APA
复制
导出至
BibTeX EndNote RefMan NoteFirst NoteExpress
×
提示
您的信息不完整,为了账户安全,请先补充。
现在去补充
×
提示
您因"违规操作"
具体请查看互助需知
我知道了
×
提示
确定
请完成安全验证×
copy
已复制链接
快去分享给好友吧!
我知道了
右上角分享
点击右上角分享
0
联系我们:info@booksci.cn Book学术提供免费学术资源搜索服务,方便国内外学者检索中英文文献。致力于提供最便捷和优质的服务体验。 Copyright © 2023 布克学术 All rights reserved.
京ICP备2023020795号-1
ghs 京公网安备 11010802042870号
Book学术文献互助
Book学术文献互助群
群 号:481959085
Book学术官方微信