基于深度学习的公路货车车型识别

张念; 张亮

doi:10.19818/j.cnki.1671-1637.2023.01.020

基于深度学习的公路货车车型识别

doi: 10.19818/j.cnki.1671-1637.2023.01.020

张念^{1, 2,},
张亮²

1.
太原理工大学土木工程学院，山西太原 030024
2.
太原科技大学交通与物流学院，山西太原 030024

基金项目:

国家自然科学基金项目 52178341

山西省回国留学人员科研项目 2020038

详细信息

作者简介:
张念(1984-)，男，湖北襄阳人，太原理工大学副教授，工学博士，从事算法优化与工程安全研究

中图分类号: U495
计量
- 文章访问数: 537
- HTML全文浏览量: 173
- PDF下载量: 153
- 被引次数: 0
出版历程
- 收稿日期: 2022-07-31
- 网络出版日期: 2023-03-08
- 刊出日期: 2023-02-25

Type recognition of highway trucks based on deep learning

ZHANG Nian^{1, 2
,},
ZHANG Liang²

1.
College of Civil Engineering, Taiyuan University of Technology, Taiyuan 030024, Shanxi, China
2.
School of Transportation and Logistics, Taiyuan University of Science and Technology, Taiyuan 030024, Shanxi, China

Funds:

National Natural Science Foundation of China 52178341

Research Project of Shanxi Scholarship Council 2020038

More Information

Author Bio:
ZHANG Nian(1984-), male, associate professor, PhD, zhangnian@tyut.edu.cn

摘要

摘要: 为判断公路货车车型，并提升货车车型识别的速度与精度，提出基于深度学习的方法对公路货车及其轮轴进行精细化目标检测；采用道路监控拍摄和网络爬取的方式获得了16 403张公路货车侧方图像，建立了货车侧方图像数据集，并采用Retinex理论和加入限制对比度的自适应直方图均衡化(CLAHE)等视觉增强方法预处理所采集图像中的光照不均图像和夜视图像；通过理论分析和对比试验选取单阶段检测网络YOLOv3作为公路货车车型识别的目标检测网络，并从调整先验框和模型输入大小以及引入注意力机制3个方面优化了检测模型；针对单帧图像可能同时出现多辆货车的情况，采用基于目标位置信息挖掘的算法分析了货车与轮轴的位置信息，提出一种通过轮轴中心点与货车预测框位置信息判定公路货车与轮轴隶属关系的方法。研究结果表明：图像经过预处理可显著增强车辆的特征信息，优化后检测模型的网络性能得到提高，通过对目标位置信息的挖掘与利用可以很好地解决货车车型判定问题；优化后的检测模型实时检测速度可达47帧·s^-1，对公路货车车型的识别综合准确率达到了94.4%。该方法实现了对公路货车车型的无接触、快速和准确识别，为公路货车车型识别提供了新的手段，符合智慧交通系统的建设需要，可进一步提升道路服务水平。
- 交通信息工程 /
- 公路货车 /
- 深度学习 /
- 车型识别 /
- 注意力机制 /
- 图像处理
Abstract: A deep learning-based fine target detection method for highway trucks and their wheel axles was proposed to determine the vehicle types of highway trucks and improve the recognization speed and accuracy of vehicle types of trucks. 16 403 side images of trucks were obtained by the road monitoring and network crawling to build a dataset of side images of trucks. The Retinex theory and visual enhancement methods, such as the contrast limited adaptive histogram equalization (CLAHE), were used to preprocess the uneven light images and night vision ones in the collected images. Theoretical analysis and comparative experiments were conducted. The one-stage detection network YOLOv3 was selected as the target detection network for the vehicle type recognition of highway trucks. Then, the detection model was optimized from three aspects, such as adjusting the sizes of prior box and model input and introducing the attention mechanism. For the case that multiple trucks might appear at the same time in a single image, an algorithm based on the mining of target location information was employed to analyze the truck and wheel axle location information. A method was proposed to determine the subordinate relationships between the highway truck and the wheel axle according to the location information of axle center points and truck prediction boxes. Research results show that the vehicle feature information can be enhanced significantly by the image preprocessing. The network performance of the detection model improves after the optimization. The issue of determining the vehicle types of trucks can be well solved by mining and leveraging the target location information. A real-time detection speed of the optimized detection model reaches 47 frames per second. The comprehensive accuracy in recognizing the vehicle types of highway trucks is 94.4%. The method realizes the non-contact, fast, and accurate recognition of vehicle types of highway trucks, provides a new means for the vehicle type recognition of highway trucks. Meeting the construction needs of intelligent traffic systems, the proposed method can be applied to further raise the road service level.
- traffic information engineering /
- highway truck /
- deep learning /
- vehicle type recognition /
- attention mechanism /
- image processing

HTML全文

图 1 公路货车车型识别流程

Figure 1. Process of highway truck vehicle type recognition

下载: 全尺寸图片幻灯片

图 2 公路货车图像数据集

Figure 2. Highway truck image data set

下载: 全尺寸图片幻灯片

图 3 光照不均图像预处理前后对比

Figure 3. Comparison of uneven light images before and after preprocessing

下载: 全尺寸图片幻灯片

图 4 公路货车夜视图像和灰度级

Figure 4. Night vision image of highway truck and gray scales

下载: 全尺寸图片幻灯片

图 5 HE处理后的图像和灰度级

Figure 5. Image and gray scales after HE processing

下载: 全尺寸图片幻灯片

图 6 AHE处理后的图像和灰度级

Figure 6. Image and gray scales after AHE processing

下载: 全尺寸图片幻灯片

图 7 CLAHE处理后的图像和灰度级

Figure 7. Image and gray scales after CLAHE processing

下载: 全尺寸图片幻灯片

图 8 Faster RCNN和YOLOv3识别效果对比

Figure 8. Recognition effect comparison of Fast RCNN and YOLOv3

下载: 全尺寸图片幻灯片

图 9 引入注意力机制模块的网络结构

Figure 9. Network structure with attention mechanism module

下载: 全尺寸图片幻灯片

图 10 损失下降曲线

Figure 10. Decline curve of loss

下载: 全尺寸图片幻灯片

图 11 一帧图像中含有多辆货车

Figure 11. Multiple trucks in one image

下载: 全尺寸图片幻灯片

图 12 货车与轮轴隶属关系判定算法

Figure 12. Algorithm for determining subordination relationship between truck and wheel axle

下载: 全尺寸图片幻灯片

图 13 货车与轮轴隶属关系测试结果

Figure 13. Test results of subordinate relationship between truck and wheel axle

下载: 全尺寸图片幻灯片

图 14 公路货车及其轮轴动态检测识别效果

Figure 14. Dynamic detection and recognition effects of highway trucks and axles

下载: 全尺寸图片幻灯片

表 1 测试结果

Table 1. Test results

算法框架	主干网络	货车识别平均精度/%	轮轴识别平均精度/%	类平均精度均值/%	检测速度/(帧·s ^-1)
Faster RCNN	ResNet50	85.5	80.0	82.8	5.3
YOLOv3	Darknet53	83.9	74.5	79.2	56.0

下载: 导出CSV

表 2 优化网络与原网络测试结果对比

Table 2. Comparison of test results between optimized network and original network

算法框架	主干网络	货车识别平均精度/%	轮轴识别平均精度/%	类平均精度均值/%	检测速度/(帧·s^-1)
原网络	Darknet53	83.9	74.5	79.2	56
优化网络	Darknet53	97.5	91.3	94.4	47

下载: 导出CSV

参考文献(31)

[1]	王玥. 基于多特征融合的监控视频车辆检索技术研究[D]. 南京: 南京邮电大学, 2019. WANG Yue. Research on vehicle retrieval technology in monitoring video based on multi-feature fusion[D]. Nanjing: Nanjing University of Posts and Telecommunications, 2019. (in Chinese)
[2]	谭挺, 王维锋, 万剑, 等. 基于多路地感线圈的交通流量检测系统研究[J]. 华东交通大学学报, 2017, 34(2): 60-65. https://www.cnki.com.cn/Article/CJFDTOTAL-HDJT201702008.htm TAN Ting, WANG Wei-feng, WAN Jian, et al. Research of traffic flow detecting system based on multiple loop detectors[J]. Journal of East China Jiaotong University, 2017, 34(2): 60-65. (in Chinese) https://www.cnki.com.cn/Article/CJFDTOTAL-HDJT201702008.htm
[3]	张琪, 胡斌, 陈乐, 等. 轴类超声检测的波束扩散成像[J]. 应用声学, 2020, 39(4): 570-580. https://www.cnki.com.cn/Article/CJFDTOTAL-YYSN202004012.htm ZHANG Qi, HU Bin, CHEN Le, et al. Image method of shaft ultrasonic detection based on beam diffusion[J]. Journal of Applied Acoustics, 2020, 39(4): 570-580. (in Chinese) https://www.cnki.com.cn/Article/CJFDTOTAL-YYSN202004012.htm
[4]	张晓娟, 邹存名. 基于激光传感技术的车位自动检测和定位研究[J]. 激光杂志, 2020, 41(11): 37-41. https://www.cnki.com.cn/Article/CJFDTOTAL-JGZZ202011008.htm ZHANG Xiao-juan, ZOU Cun-ming. Research on automatic detection and location of parking space based on laser sensing technology[J]. Laser Journal, 2020, 41(11): 37-41. (in Chinese) https://www.cnki.com.cn/Article/CJFDTOTAL-JGZZ202011008.htm
[5]	KIM J, YOU K, CHOI H. Post-voltage-boost circuit-supported single-ended class-B amplifier for piezoelectric transducer applications[J]. Sensors, 2020, 20(18): 5412. doi: 10.3390/s20185412
[6]	PAPAGEORGIOU C P, OREN M, POGGIO T. A general framework for object detection[C]//PAPAGEORGIOU C P. Sixth IEEE International Conference on Computer Vision. New York: IEEE, 1998: 555-562.
[7]	LIENHART R, MAYDT J. An extended set of haar-like features for rapid object detection[C]//LIENHART R. IEEE International Conference on Image Processing. New York: IEEE, 2002: 900-903.
[8]	VIOLA P, JONES M J. Robust real-time face detection[J]. International Journal of Computer Vision, 2004, 57(2): 137-154. doi: 10.1023/B:VISI.0000013087.49260.fb
[9]	LOWE D G. Distinctive image features from scale-invariant keypoints[J]. International Journal of Computer Vision, 2004, 60(2): 91-110. doi: 10.1023/B:VISI.0000029664.99615.94
[10]	FELZENSZWALB P F, GIRSHICK R B, MCALLESTER D, et al. Object detection with discriminatively trained part-based models[J]. IEEE Transactions on Pattern Analysis and Machine Intelligence, 2010, 32(9): 1627-1645. doi: 10.1109/TPAMI.2009.167
[11]	DALAL S, VISHWAKARMA V P, KUMAR S. Feature- based sketch-photo matching for face recognition[J]. Procedia Computer Science, 2020, 167: 562-570. doi: 10.1016/j.procs.2020.03.318
[12]	耿庆田, 于繁华, 王宇婷, 等. 基于特征融合的车型检测新算法[J]. 吉林大学学报(工学版), 2018, 48(3): 929-935. https://www.cnki.com.cn/Article/CJFDTOTAL-JLGY201803037.htm GENG Qing-tian, YU Fan-hua, WANY Yu-ting, et al. New algorithm for vehicle type detection based on feature fusion[J]. Journal of Jilin University (Engineering and Technology Edition), 2018, 48(3): 929-935. (in Chinese) https://www.cnki.com.cn/Article/CJFDTOTAL-JLGY201803037.htm
[13]	裴月玲, 张涛. 基于人工智能的车牌号与车型识别系统[J]. 公路, 2019, 64(8): 277-281. https://www.cnki.com.cn/Article/CJFDTOTAL-GLGL201908055.htm PEI Yue-ling, ZHANG Tao. Vehicle number plate and vehicle identification system based on artificial intelligence[J]. Highway, 2019, 64(8): 277-281. (in Chinese) https://www.cnki.com.cn/Article/CJFDTOTAL-GLGL201908055.htm
[14]	李国良, 周煊赫, 孙佶, 等. 基于机器学习的数据库技术综述[J]. 计算机学报, 2020, 43(11): 2019-2049. https://www.cnki.com.cn/Article/CJFDTOTAL-JSJX202011001.htm LI Guo-liang, ZHOU Xuan-he, SUN Ji, et al. A survey of machine learning based database techniques[J]. Chinese Journal of Computers, 2020, 43(11): 2019-2049. (in Chinese) https://www.cnki.com.cn/Article/CJFDTOTAL-JSJX202011001.htm
[15]	李珣, 刘瑶, 李鹏飞, 等. 基于Darknet框架下YOLOv2算法的车辆多目标检测方法[J]. 交通运输工程学报, 2018, 18(6): 142-158. doi: 10.19818/j.cnki.1671-1637.2018.06.015 LI Xun, LIU Yao, LI Peng-fei, et al. Vehicle multi-target detection method based on YOLOv2 algorithm under Darknet framework[J]. Journal of Traffic and Transportation Engineering, 2018, 18(6): 142-158. (in Chinese) doi: 10.19818/j.cnki.1671-1637.2018.06.015
[16]	黎哲明, 蔡鸿明, 姜丽红. 基于卷积神经网络的车辆品牌和型号识别[J]. 东华大学学报(自然科学版), 2017, 43(4): 472-477. https://www.cnki.com.cn/Article/CJFDTOTAL-DHDZ201704003.htm LI Zhe-ming, CAI Hong-ming, JIANG Li-hong. Car make and model recognition based on convolutional neural network[J]. Journal of Donghua University(Natural Science), 2017, 43(4): 472-477. (in Chinese) https://www.cnki.com.cn/Article/CJFDTOTAL-DHDZ201704003.htm
[17]	袁公萍, 汤一平, 韩旺明, 等. 基于深度卷积神经网络的车型识别方法[J]. 浙江大学学报(工学版), 2018, 52(4): 694-702. https://www.cnki.com.cn/Article/CJFDTOTAL-ZDZC201804012.htm YUAN Gong-ping, TANG Yi-ping, HAN Wang-ming, et al. Vehicle category recognition based on deep convolutional neural network[J]. Journal of Zhejiang University (Engineering Science), 2018, 52(4): 694-702. (in Chinese) https://www.cnki.com.cn/Article/CJFDTOTAL-ZDZC201804012.htm
[18]	陈立潮, 张雷, 曹建芳, 等. 梯度直方图卷积特征的胶囊网络在交通监控下的车型分类[J]. 计算机应用, 2020, 40(10): 2881-2889. https://www.cnki.com.cn/Article/CJFDTOTAL-JSJY202010015.htm CHEN Li-chao, ZHANG Lei, CAO Jian-fang, et al. Vehicle classification based on HOG-C CapsNet in traffic surveillance scenarios[J]. Journal of Computer Applications, 2020, 40(10): 2881-2889. (in Chinese) https://www.cnki.com.cn/Article/CJFDTOTAL-JSJY202010015.htm
[19]	陈树东. 基于卷积神经网络的车辆检测与车型辨识算法研究[D]. 成都: 电子科技大学, 2018. CHEN Shu-dong. Vehicle detection and vehicle type identification algorithm based on convolutional neural network[D]. Chengdu: University of Electronic Science and Technology of China, 2018. (in Chinese)
[20]	贺甜. 基于神经网络的货车车型识别系统研究[D]. 西安: 西安工业大学, 2019. HE Tian. Research on truck model recognition system based on neural network[D]. Xi'an: Xi'an Technological University, 2019. (in Chinese)
[21]	黄涛. 压电薄膜传感器在自动车型分类器的应用[J]. 传感器世界, 2017, 23(7): 21-24. https://www.cnki.com.cn/Article/CJFDTOTAL-CGSJ201707003.htm HUANG Tao. Application of piezoelectric thin film sensor in automatic vehicle classification[J]. Sensor World, 2017, 23(7): 21-24. (in Chinese) https://www.cnki.com.cn/Article/CJFDTOTAL-CGSJ201707003.htm
[22]	夏发钦. 利用地感应线圈检测机动车辆的原理与实现[D]. 武汉: 武汉科技大学, 2011. XIA Fa-qin. The principle and realization of using ground induction coils detection motor vehicles[D]. Wuhan: Wuhan University of Science and Technology, 2011. (in Chinese)
[23]	郭敏. 基于高速公路大货车违法占道监测系统的车辆检测与跟踪研究[D]. 成都: 西南交通大学, 2015. GUO Min. Research on vehicle detection and tracking based on the large trucks illegal lane highway monitoring system[D]. Chengdu: Southwest Jiaotong University, 2015. (in Chinese)
[24]	周俊宇, 赵艳明. 卷积神经网络在图像分类和目标检测应用综述[J]. 计算机工程与应用, 2017, 53(13): 34-41. https://www.cnki.com.cn/Article/CJFDTOTAL-JSGG201713006.htm ZHOU Jun-yu, ZHAO Yan-ming. Application of convolution neural network in image classification and object detection[J]. Computer Engineering and Applications, 2017, 53(13): 34-41. (in Chinese) https://www.cnki.com.cn/Article/CJFDTOTAL-JSGG201713006.htm
[25]	JOBSON D J, RAHMAN Z, WOODELL G A. A multiscale Retinex for bridging the gap between color images and the human observation of scenes[J]. IEEE Transactions on Image Processing, 1997, 6(7): 965-976.
[26]	FUNT B, CIUREA F, MCCANN J. Retinex in MATLAB^TM[J]. Journal of Electronic Imaging, 2004, 13(1): 112-121.
[27]	MAJEED S H, ISA N A M. Adaptive entropy index histogram equalization for poor contrast images[J]. IEEE Access, 2020, 9: 6402-6437.
[28]	SANTOS J C M, CARRIJO G A, FÁTIMA DOS SANTOS CARDOSO C, et al. Fundus image quality enhancement for blood vessel detection via a neural network using CLAHE and Wiener filter[J]. Research on Biomedical Engineering, 2020, 36(2): 107-117.
[29]	桑军, 郭沛, 项志立, 等. Faster-RCNN的车型识别分析[J]. 重庆大学学报, 2017, 40(7): 32-36. https://www.cnki.com.cn/Article/CJFDTOTAL-FIVE201707005.htm SANG Jun, GUO Pei, XIANG Zhi-li, et al. Vehicle detection based on faster-RCNN[J]. Journal of Chongqing University, 2017, 40(7): 32-36. (in Chinese) https://www.cnki.com.cn/Article/CJFDTOTAL-FIVE201707005.htm
[30]	鞠默然, 罗海波, 王仲博, 等. 改进的YOLO v3算法及其在小目标检测中的应用[J]. 光学学报, 2019, 39(7): 0715004. https://www.cnki.com.cn/Article/CJFDTOTAL-GXXB201907028.htm JU Mo-ran, LUO Hai-bo, WANG Zhong-bo, et al. Improved YOLO v3 algorithm and its application in small target detection[J]. Acta Optica Sinica, 2019, 39(7): 0715004. (in Chinese) https://www.cnki.com.cn/Article/CJFDTOTAL-GXXB201907028.htm
[31]	张如意. 基于深度学习的视频目标检测算法研究[D]. 北京: 北京交通大学, 2020. ZHANG Ru-yi. Video object detection based on deep learning[D]. Beijing: Beijing Jiaotong University, 2020. (in Chinese)