面向高速公路的车辆视频监控分析系统

doi:10.19665/j.issn1001-2400.2021.05.021

摘要/Abstract

摘要：

随着视频监控技术在道路安全应用的迅猛发展,为实现高速公路智能化管理,提出一套面向高速公路的车辆视频监控分析系统。通过对监控视频流中的车辆进行检测和跟踪,进一步实现了高速公路相关车辆监测的应用。提出了基于双向金字塔多尺度融合的轻量级车辆检测跟踪算法,基于YOLOv3在主干网络上使用轻量级网络EfficientNet,并且利用双向特征金字塔网络进行多尺度特征融合,使得算法在保证检测实时性的同时提升检测的准确度。通过采集高速公路监测视频,构建了一个多场景高速公路车辆目标数据集。在此数据集上的实验结果表明,所提出的算法检测精度达97.11%,高于原始YOLOv3检测算法16.5%,并且结合DeepSORT模型在车辆跟踪上以31帧每秒的帧速度实时运行。同时,该车辆监测系统可在车流量统计、交通异常事件检测领域进行多路实时监测,具有实际应用价值。

关键词: 高速公路视频监控, 车辆监测, 目标检测, 目标跟踪, 多尺度特征融合

Abstract:

With the rapid development of video surveillance technology in the application of road safety,in order to realize the intelligent management of the expressway,this paper proposes a vehicle video surveillance and analysis system for the expressway.By detecting and tracking the vehicles for the surveillance videos,the applications of expressway related vehicle monitoring are further realized.The system presents a lightweight vehicle detection and tracking algorithm based on bidirectional pyramid multi-scale integration.The algorithm uses the lightweight network EfficientNet based on YOLOv3,and uses the bidirectional feature pyramid network (BiFPN) for multi-scale feature fusion.This system could ensure the real-time detection and improve the detection accuracy.Furthermore,in this paper,a multi-scene-highway-vehicles dataset is constructed by collecting freeway monitoring videos.Experimental results of this dataset shows that the detection accuracy of the proposed algorithm is 97.11%,which is 16.5% higher than that of the original YOLOv3 detection algorithm,and that the algorithm could run in real time at 31fps on vehicle tracking by combining with the DeepSORT model.At the same time,the vehicle monitoring system could realize multi-channel real-time detections in the field of vehicle flow statistics and traffic abnormal event detection,which is of practical application value.

Key words: highway video surveillance, vehicle monitoring, object detection, object tracking, multi-scale feature fusion

中图分类号:

TP23

毛昭勇,王亦晨,王鑫,沈钧戈. 面向高速公路的车辆视频监控分析系统[J]. 西安电子科技大学学报, 2021, 48(5): 178-189.

MAO Zhaoyong,WANG Yichen,WANG Xin,SHEN Junge. Vehicle video surveillance and analysis system for the expressway[J]. Journal of Xidian University, 2021, 48(5): 178-189.

图/表 17

图1

图2

图3

图4

图5

图6

图7

图8

图9

表1

表2

表3

图10

图11

表4

图12

表5

参考文献 16

[1]	ALMOMANI M, ALKHALIL N Y, AHMAD E M, et al. Ubiquitous GPS Vehicle Tracking and Management System[C]// 2011 IEEE Jordan Conference on Applied Electrical Engineering and Computing Technologies.Piscataway:IEEE, 2011:1-6.
[2]	SHRUTHI S. Vehicle Tracking Using Convolutional Neural Network[J]. Lecture Notes in Engineering & Computer Science, 2011:1052-1055.
[3]	LIANG Y, LI Y, ZHAO K, et al. Object Tracking Algorithm Based on Multi-Channel Extraction of AHLBP Texture Features[C]// 2018 International Conference on Advanced Mechatronic Systems.Piscataway:IEEE, 2018,332-336.
[4]	YANG Y, MU C, YANG R, et al. Real-Time Vehicle Tracking Using Convolutional Neural Networks in Aerial Video[C]// 2019 International Conference on Sensing,Diagnostics,Prognostics,and Control.Piscataway:IEEE, 2019:251-255.
[5]	GIRSHICK R, DONAHUE J, DARRELL T, et al. Rich Feature Hierarchies for Accurate Object Detection and Semantic Segmentation[C]// Proceedings of the IEEE Conference on Computer Vision and Pattern Recognition.Piscataway:IEEE, 2014:580-587.
[6]	GIRSHICK R. Fast R-Cnn[C]// Proceedings of the IEEE International Conference on Computer Vision.Piscataway:IEEE, 2015:1440-1448.
[7]	REN S, HE K, GIRSHICK R, et al. Faster R-Cnn:Towards Real-Time Object Detection With Region Proposal Networks[J]. IEEE Transactions on Pattern Analysis and Machine Intellingemce, 2017, 39(8):1137-1149.
[8]	REDMON J, DIVVALA S, GIRSHICK R, et al. You Only Look Once:Unified,Real-Time Object Detection[C]// 2016 IEEE Conference on Computer Vision and Pattern Recognition.Piscataway:IEEE, 2016:779-788.
[9]	REDMON J, FARHADI A. Yolov3:An Incremental Improvement[C/OL]. [2021-04-30]. https://arxiv.org/cs/arxiv:1804.02767.
[10]	TAN M, LE Q. EfficientNet:Rethinking Model Scaling for Convolutional Neural Networks[C]// 2019 International Conference on Machine Learning.Long Beach:PMLR, 2019:6105-6114.
[11]	LIN T, DOLLAR P, GIRSHICK R, et al. Feature Pyramid Networks for Object Detection[C]// 2017 IEEE Conference on Computer Vision and Pattern Recognition.Piscataway:IEEE, 2017:936-944.
[12]	LIU S, QI L, QIN H, et al. Path Aggregation Network for Instance Segmentation[C]// 2018 IEEE/CVF Conference on Computer Vision and Pattern Recognition.Piscataway:IEEE, 2018:8759-8768.
[13]	BEWLEY A, GE Z, OTT L, et al. Simple Online and Realtime Tracking[C]// 2016 IEEE International Conference on Image Processing.Piscataway:IEEE, 2016:3464-3468.
[14]	WOJKE N, BEWLEY A, PAULUS D. Simple Online and Realtime Tracking with A Deep Association Metric[C]// 2017 IEEE International Conference on Image Processing.Piscataway:IEEE, 2017:3645-3649.
[15]	SANDLER M, HOWARD A, ZHU M, et al. MobileNetV2:Inverted Residuals and Linear Bottlenecks[C]// 2018 IEEE/CVF Conference on Computer Vision and Pattern Recognition.Piscataway:IEEE, 2018:4510-4520.
[16]	TAN M, PANG R, LE Q V. Efficientdet:Scalable and Efficient Object Detection[C]// 2020 Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition.Piscataway:IEEE, 2020:10781-10790.

真实情况	预测结果
真实情况	正例(Positive)	反例(Negative)
真(True)	TP(真正例)	FN(真反例)
假(False)	FP(假正例)	TN(假反例)

实验编号	检测模型	模型参数量大小/MB	mAP/%	平均耗时/ms
1	YOLOv3	246.4	83.37	17
2	YOLOv3-tiny	34.7	76.05	6
3	YOLOv3-EffiNet	62.9	91.47	11
4	YOLOv3-EffiNet+BiFPN	67.6	97.11	18

检测模型	YOLOv3	YOLOv3-tiny	YOLOv3-EffiNet	YOLOv3-EffiNet+BiFPN
小汽车	92.35	83.51	98.03	98.73
货车	86.08	78.24	99.84	98.09
出租车	78.13	72.62	79.34	97.75
面包货车	76.94	69.81	88.65	93.85

检测模型	上行车辆/辆	下行车辆/辆	总车辆数/辆	准确率/%	平均帧速度/(帧/秒)
YOLOv3+DeepSORT	692	521	1 213	81.9	31
YOLOv3-tiny+DeepSORT	427	339	766	51.7	142
YOLOv3-EffiNet+DeepSORT	718	557	1 275	86.1	76
YOLOv3-EffiNet+BiFPN+DeepSORT	792	590	1 382	93.3	31

异常事件种类	真实数量	错检数量	漏检数量	检测数量	准确率/%	平均帧速度 /(帧/秒)
异常停车	98段	0段	7段	91段	92.9	29
拥堵	117段	0段	4段	113段	96.6	29
逆行	57段	1段	4段	53段	91.2	28