Lightweight Real-Time Object Detection via Enhanced Global Perception and Intra-Layer Interaction for Complex Traffic Scenarios

IF 1.3 4区 工程技术 Q3 COMPUTER SCIENCE, INFORMATION SYSTEMS
ben liang;Jia Su;Kangkang Feng;Yongqiang Zhang;Weimin Hou
{"title":"Lightweight Real-Time Object Detection via Enhanced Global Perception and Intra-Layer Interaction for Complex Traffic Scenarios","authors":"ben liang;Jia Su;Kangkang Feng;Yongqiang Zhang;Weimin Hou","doi":"10.1109/TLA.2024.10472963","DOIUrl":null,"url":null,"abstract":"Due to unfavorable factors such as cluttered spatial and temporal distribution of multiple types of targets, occlusion of background objects of different shapes, and blurring of feature information by inclement weather, the low detection accuracy in complex traffic scenarios has been a troubling issue. Regarding the above-mentioned issues, the paper proposes a lightweight real-time detection network to augment multi-scale object perception capabilities in traffic scenarios while ensuring real-time detection speed. First, we construct a novel global feature extraction (GFE) structure by cascading orthogonal band convolution kernels that capture the global dependencies between pixels to improve feature discrimination. Then, an intra-layer multi-scale feature interaction (IMFI) module is proposed to reinforce the effective reuse and multi-level transfer of salient features. In addition, we build a multi-branch scale-aware aggregation (MSA) module that captures abundant context-associated features to improve the target decision-making capability and the self-adaptive capability of the model when dealing with diverse object scales. Experimental results demonstrate that the proposed approach attains a significant improvement of 5.6 percentage points in AP50 with fewer parameters and computational power compared to the baseline model, with an improved FPS of 73. Furthermore, our approach strikes the optimal speed-accuracy balance when compared against other excellent object detection algorithms of the same magnitude.","PeriodicalId":55024,"journal":{"name":"IEEE Latin America Transactions","volume":null,"pages":null},"PeriodicalIF":1.3000,"publicationDate":"2024-03-18","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"https://ieeexplore.ieee.org/stamp/stamp.jsp?tp=&arnumber=10472963","citationCount":"0","resultStr":null,"platform":"Semanticscholar","paperid":null,"PeriodicalName":"IEEE Latin America Transactions","FirstCategoryId":"5","ListUrlMain":"https://ieeexplore.ieee.org/document/10472963/","RegionNum":4,"RegionCategory":"工程技术","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":null,"EPubDate":"","PubModel":"","JCR":"Q3","JCRName":"COMPUTER SCIENCE, INFORMATION SYSTEMS","Score":null,"Total":0}
引用次数: 0

Abstract

Due to unfavorable factors such as cluttered spatial and temporal distribution of multiple types of targets, occlusion of background objects of different shapes, and blurring of feature information by inclement weather, the low detection accuracy in complex traffic scenarios has been a troubling issue. Regarding the above-mentioned issues, the paper proposes a lightweight real-time detection network to augment multi-scale object perception capabilities in traffic scenarios while ensuring real-time detection speed. First, we construct a novel global feature extraction (GFE) structure by cascading orthogonal band convolution kernels that capture the global dependencies between pixels to improve feature discrimination. Then, an intra-layer multi-scale feature interaction (IMFI) module is proposed to reinforce the effective reuse and multi-level transfer of salient features. In addition, we build a multi-branch scale-aware aggregation (MSA) module that captures abundant context-associated features to improve the target decision-making capability and the self-adaptive capability of the model when dealing with diverse object scales. Experimental results demonstrate that the proposed approach attains a significant improvement of 5.6 percentage points in AP50 with fewer parameters and computational power compared to the baseline model, with an improved FPS of 73. Furthermore, our approach strikes the optimal speed-accuracy balance when compared against other excellent object detection algorithms of the same magnitude.
通过增强的全局感知和层内交互进行轻量级实时物体检测,以应对复杂的交通场景
由于多种类型目标的时空分布杂乱、不同形状的背景物体遮挡、恶劣天气对特征信息的模糊等不利因素,复杂交通场景下的低检测精度一直是个令人头疼的问题。针对上述问题,本文提出了一种轻量级实时检测网络,以增强交通场景中的多尺度物体感知能力,同时保证实时检测速度。首先,我们通过级联正交频带卷积核构建了新颖的全局特征提取(GFE)结构,捕捉像素之间的全局依赖关系以提高特征辨别能力。然后,我们提出了层内多尺度特征交互(IMFI)模块,以加强突出特征的有效重用和多层次转移。此外,我们还建立了一个多分支尺度感知聚合(MSA)模块,捕捉丰富的上下文相关特征,以提高目标决策能力和模型在处理不同物体尺度时的自适应能力。实验结果表明,与基线模型相比,所提出的方法以更少的参数和计算能力将 AP50 显著提高了 5.6 个百分点,FPS 提高了 73。此外,与其他同级别的优秀物体检测算法相比,我们的方法在速度和准确性之间取得了最佳平衡。
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
IEEE Latin America Transactions
IEEE Latin America Transactions COMPUTER SCIENCE, INFORMATION SYSTEMS-ENGINEERING, ELECTRICAL & ELECTRONIC
CiteScore
3.50
自引率
7.70%
发文量
192
审稿时长
3-8 weeks
期刊介绍: IEEE Latin America Transactions (IEEE LATAM) is an interdisciplinary journal focused on the dissemination of original and quality research papers / review articles in Spanish and Portuguese of emerging topics in three main areas: Computing, Electric Energy and Electronics. Some of the sub-areas of the journal are, but not limited to: Automatic control, communications, instrumentation, artificial intelligence, power and industrial electronics, fault diagnosis and detection, transportation electrification, internet of things, electrical machines, circuits and systems, biomedicine and biomedical / haptic applications, secure communications, robotics, sensors and actuators, computer networks, smart grids, among others.
×
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