基于优化卷积神经网络的车辆特征识别算法研究

doi:10.11959/j.issn.1000-0801.2023188

电信科学 ›› 2023, Vol. 39 ›› Issue (10): 101-111.doi: 10.11959/j.issn.1000-0801.2023188

• 研究与开发 • 上一篇

基于优化卷积神经网络的车辆特征识别算法研究

陈暄¹, 吴吉义²^,³

¹ 浙江工业职业技术学院，浙江绍兴 312000
² 浙江省人工智能学会，浙江杭州 310027
³ 浙江大学智能教育研究中心，浙江杭州 310027

修回日期:2023-10-10 出版日期:2023-10-01 发布日期:2023-10-01
作者简介:陈暄（1979- ），男，浙江工业职业技术学院副教授，主要研究方向为云计算、人工智能
吴吉义（1980- ），男，博士，浙江大学高级工程师，主要研究方向为服务计算、人工智能
基金资助:
国家自然科学基金资助项目(61702151);国家自然科学基金资助项目(61702320);国家自然科学基金资助项目(61772334);国家重点研发计划项目(2018YFB1003800);浙江省哲学社会科学规划课题(23NDJC369YB)

Research on vehicle feature recognition algorithm based on optimized convolutional neural network

Xuan CHEN¹, Jiyi WU²^,³

¹ Zhejiang Industry Polytechnic College, Shaoxing 312000, China
² Zhejiang Federation of Artificial Intelligence, Hangzhou 310027, China
³ Intelligent Education Research Center, Zhejiang University, Hangzhou 310027, China

Revised:2023-10-10 Online:2023-10-01 Published:2023-10-01
Supported by:
The National Natural Science Foundation of China(61702151);The National Natural Science Foundation of China(61702320);The National Natural Science Foundation of China(61772334);The National Key Research and Development Program of China(2018YFB1003800);Zhejiang Provincial Philosophy and Social Sciences Planning Project(23NDJC369YB)

摘要/Abstract

摘要：

针对道路场景图像中不同距离目标车辆特征存在识别效果弱、精度低的问题，提出一种基于优化卷积神经网络的车辆特征识别算法。首先，采用基于PAN模型的多尺度输入获取不同距离的目标车辆特征；其次，在卷积神经网络结构中加入多池化、BN层和Leaky ReLU激活函数改进网络模型的性能，通过引入混合注意力机制，集中关注车辆图像中的重要特征和区域，从而增强了网络模型的泛化能力；最后，通过构建多层次卷积神经网络结构完成对车辆的特征效果识别。仿真实验结果表明，在单一场景的BIT-Vehicle数据库中，本文算法相比CNN、R-CNN、ABC-CNN、Faster R-CNN、AlexNet、VGG16和YOLOV8在单一目标和多目标识别率方面分别提高了 16.75%、10.9%、4%、3.7%、2.46%、1.3%、1%和 17.8%、10.5%、2.5%、3.8%、2.7%、1.1%、1.3%，在复杂场景的UA-DETRAC数据库中，本文算法相比其他算法在不同距离目标车辆识别中获得了更加精确的效果。

关键词: 车辆识别, 卷积神经网络, 多尺度输入

Abstract:

To address the issue of weak identification and low accuracy in recognizing features of target vehicles at different distances in road scene images, a vehicle feature recognition algorithm based on optimized convolutional neural network (CNN) was proposed.Firstly, a multi-scale input based on the PAN model was employed to capture target vehicle features at varying distances.Subsequently, improvements were made to the network model by incorporating multi-pool, batch normalization (BN) layers, and Leaky ReLU activation functions within the CNN architecture.Furthermore, the generalization ability of the network model was enhanced by introducing a hybrid attention mechanism that focuses on important features and regions in the vehicle image.Lastly, a multi-level CNN structure was constructed to achieve feature recognition for vehicles.Simulation experiment results conducted on the BIT-Vehicle database within a single scene show the proposed algorithm’s significant enhancements in single-object and multi-object recognition rates compared to CNN, R-CNN, ABC-CNN, Faster R-CNN, AlexNet, VGG16, and YOLOV8.Specifically, improvements of 16.75%, 10.9%, 4%, 3.7%, 2.46%, 1.3%, and 1% in single-object recognition, as well as 17.8%, 10.5%, 2.5%, 3.8%, 2.7%, 1.1%, and 1.3% in multi-object recognition, have been demonstrated by the proposed algorithm, respectively.Over the more complex UA-DETRAC datasets, more precise results have been also achieved by the proposed algorithm in recognizing target vehicles at various distances compared to other algorithms.

Key words: vehicle recognition, convolutional neural network, multi-scale input

中图分类号:

TP391.14

陈暄, 吴吉义. 基于优化卷积神经网络的车辆特征识别算法研究[J]. 电信科学, 2023, 39(10): 101-111.

Xuan CHEN, Jiyi WU. Research on vehicle feature recognition algorithm based on optimized convolutional neural network[J]. Telecommunications Science, 2023, 39(10): 101-111.

图/表 12

图1

图2

图3

图4

图5

表1

表2

表3

图6

图7

图8

图9

参考文献 21

[1]	ZHANG J P , WANG F Y , WANG K F ,et al. Data-driven intelligent transportation systems:a survey[J]. IEEE Transactions on Intelligent Transportation Systems, 2011,12(4): 1624-1639.
[2]	COLLINS R T , LIPTON A J , KANADE T . Introduction to the special section on video surveillance[J]. IEEE Transactions on Pattern Analysis and Machine Intelligence, 2000,22(8): 745-746.
[3]	CHAROUH Z , EZZOUHRI A , GHOGHO M ,et al. A resource-efficient CNN-based method for moving vehicle detection[J]. Sensors, 2022,22(3): 1193.
[4]	RANJITHKUMAR S , CHENTHUR PANDIAN S . Automatic license plate recognition system for vehicles using a CNN[J]. Computers,Materials ＆ Continua, 2022,71(1): 35-50.
[5]	SHI R W , YANG S C , CHEN Y Y ,et al. CNN-Transformer for visual-tactile fusion applied in road recognition of autonomous vehicles[J]. Pattern Recognition Letters, 2023,166: 200-208.
[6]	SATYANARAYANA G S R , DESHMUKH P , DAS S K . Vehicle detection and classification with spatio-temporal information obtained from CNN[J]. Displays, 2022,75:102294.
[7]	SOON F C , KHAW H Y , CHUAH J H ,et al. Hyper-parameters optimisation of deep CNN architecture for vehicle logo recognition[J]. IET Intelligent Transport Systems, 2018,12(8): 939-946.
[8]	王上, 唐欢容 . 一种基于混合粒子群优化算法的深度卷积神经网络架构搜索方法[J]. 计算机应用研究, 2023,40(7): 2019-2024.
	WANG S , TANG H R . Deep convolutional neural architecture search method based on hybrid particle swarm optimization algorithm[J]. Application Research of Computers, 2023,40(7): 2019-2024.
[9]	KIYMA? E , KAYA Y . A novel automated CNN arrhythmia classifier with memory-enhanced artificial hummingbird algorithm[J]. Expert Systems With Applications, 2023(213): 119162.
[10]	GHASEMI DAREHNAEI Z , SHOKOUHIFAR M , YAZDANJOUEI H ,et al. SI-EDTL:swarm intelligence ensemble deep transfer learning for multiple vehicle detection in UAV images[J]. Concurrency and Computation:Practice and Experience, 2022,34(5): e6726.
[11]	MAITY M , BANERJEE S , SINHA CHAUDHURI S . Faster R-CNN and YOLO based vehicle detection:a survey[C]// Proceedings of 2021 5th International Conference on Computing Methodologies and Communication (ICCMC). Piscataway:IEEE Press, 2021: 1442-1447.
[12]	GHOSH R . On-road vehicle detection in varying weather conditions using Faster R-CNN with several region proposal networks[J]. Multimedia Tools and Applications, 2021,80(17): 25985-25999.
[13]	HSU S C , HUANG C L , CHUANG C H . Vehicle detection using simplified Fast R-CNN[C]// Proceedings of 2018 International Workshop on Advanced Image Technology (IWAIT). Piscataway:IEEE Press, 2018: 1-3.
[14]	宁俊, 王年, 朱明 . 基于改进Faster R-CNN的车辆类型识别算法[J]. 安徽大学学报(自然科学版), 2021,45(3): 26-33.
	NING J , WANG N , ZHU M . Vehicle type recognition algorithm based on the improved Faster R-CNN[J]. Journal of Anhui University (Natural Sciences), 2021,45(3): 26-33.
[15]	LUO J Q , FANG H S , SHAO F M ,et al. Multi-scale traffic vehicle detection based on Faster R-CNN with NAS optimization and feature enrichment[J]. Defence Technology, 2021,17(4): 1542-1554.
[16]	LIU S , QI L , QIN H F ,et al. Path aggregation network for instance segmentation[C]// Proceedings of 2018 IEEE/CVF Conference on Computer Vision and Pattern Recognition. Piscataway:IEEE Press, 2018: 8759-8768.
[17]	GAO X , WANG J F , ZHOU M Z . The research of resource allocation method based on GCN-LSTM in 5G network[J]. IEEE Communications Letters, 2023,27(3): 926-930.
[18]	ERKAN U , TOKTAS A , USTUN D . Hyperparameter optimization of deep CNN classifier for plant species identification using artificial bee colony algorithm[J]. Journal of Ambient Intelligence and Humanized Computing, 2023,14(7): 8827-8838.
[19]	ZHANG X L , CUI J , LIU H J ,et al. Weed identification in soybean seedling stage based on optimized Faster R-CNN algorithm[J]. Agriculture, 2023,13(1): 175.
[20]	DONG Z , WU Y W , PEI M T ,et al. Vehicle type classification using a semisupervised convolutional neural network[J]. IEEE Transactions on Intelligent Transportation Systems, 2015,16(4): 2247-2256.
[21]	WEN L Y , DU D W , CAI Z W ,et al. UA-DETRAC:a new benchmark and protocol for multi-object detection and tracking[J]. Computer Vision and Image Understanding, 2020(193): 102907.

算法	sedan	SUV	truck	bus	minivan	microbus
本文算法	96.32%	95.61%	92.14%	94.32%	93.89%	94.78%
CNN	78.38%	82.41%	79.23%	81.92%	82.14%	81.62%
R-CNN	84.28%	87.29%	81.82%	85.23%	86.24%	86.19%
ABC-CNN	90.73%	89.31%	88.29%	93.02%	91.21%	92.18%
Faster R-CNN	91.21%	88.16%	89.27%	93.32%	91.82%	92.91%
AlexNet	92.78%	91.40%	90.75%	93.46%	92.04%	93.01%
VGG16	94.19%	93.39%	91.74%	93.52%	92.76%	93.85%
YOLOV8	95.12%	94.02%	91.84%	93.75%	92.79%	93.92%

算法	sedan	SUV	truck	bus	minivan	microbus
本文算法	92.13%	90.13%	90.23%	91.32%	90.41%	92.38%
CNN	76.18%	79.31%	75.73%	78.93%	78.34%	75.32%
R-CNN	81.72%	85.17%	78.92%	82.43%	83.14%	83.27%
ABC-CNN	89.34%	88.48%	87.91%	89.03%	88.71%	89.31%
Faster R-CNN	88.31%	89.27%	83.71%	88.92%	87.31%	88.93%
AlexNet	90.78%	89.40%	84.75%	89.47%	88.34%	89.17%
VGG16	91.24%	89.42%	87.32%	90.43%	89.04%	91.05%
YOLOV8	91.39%	90.01%	88.19%	90.47%	89.13%	91.38%

算法	sedan	SUV	truck	bus	minivan	microbus
本文算法	12.183	13.238	11.147	11.611	12.523	12.416
CNN	15.982	15.298	14.327	14.317	15.925	15.291
R-CNN	14.925	14.198	13.782	12.893	14.197	14.981
ABC-CNN	13.891	13.278	12.982	12.892	13.183	13.261
Faster R-CNN	13.782	14.372	12.931	12.927	13.894	13.936
AlexNet	13.153	14.176	12.369	12.725	13.126	13.619
VGG16	13.013	13.739	12.319	12.074	13.015	12.946
YOLOV8	12.186	13.341	11.159	11.657	12.576	12.519

基于优化卷积神经网络的车辆特征识别算法研究

Research on vehicle feature recognition algorithm based on optimized convolutional neural network

在线阅读

PDF下载

可视化

摘要/Abstract

引用本文

使用本文

图/表 12

参考文献 21

相关文章 10

Metrics

推荐阅读 0

[1]	陶涛, 李珍, 王冀彬, 徐海勇, 江勇, 陈卓, 张润波, 胡清源. 基于图神经网络的权益推荐技术方案研究[J]. 电信科学, 2023, 39(8): 91-101.
[2]	云本胜, 干潇雅, 钱亚冠. 一种基于随机森林和改进卷积神经网络的网络流量分类方法[J]. 电信科学, 2023, 39(7): 80-89.
[3]	陈靓, 钱亚冠, 何志强, 关晓惠, 王滨, 王星. 深度卷积神经网络的柔性剪枝策略[J]. 电信科学, 2022, 38(1): 83-94.
[4]	刘子豪, 贾小军, 张素兰, 徐志玲, 张俊. 一种融合MeanShift聚类分析和卷积神经网络的Vibe++背景分割方法[J]. 电信科学, 2021, 37(3): 133-145.
[5]	曹靖城, 张继东, 史国杰. 一种使用边缘增强技术提高相似图片检索召回率的方法[J]. 电信科学, 2021, 37(1): 76-84.
[6]	贾川民, 赵政辉, 王苫社, 马思伟. 基于神经网络的图像视频编码[J]. 电信科学, 2019, 35(5): 32-42.
[7]	苏赋,吕沁,罗仁泽. 基于深度学习的图像分类研究综述[J]. 电信科学, 2019, 35(11): 58-74.
[8]	李强,康子路. 基于深度时空特征卷积—池化的视频人群计数方法[J]. 电信科学, 2018, 34(6): 72-79.
[9]	徐宏伟,严迪群,阳帆,王让定,金超,向立. 基于卷积神经网络的电子变调语音检测算法[J]. 电信科学, 2018, 34(2): 46-57.
[10]	邢惠钧,昌硕. 基于移动小车的行人监控系统[J]. 电信科学, 2017, 33(2): 120-127.