随着社会经济的发展，我国机动车保有量不断攀升，交通拥堵问题愈发严重，增加了人们的通勤时间，提高了交通事故发生率。相较于现有固定配时，本文提出了一种基于机器视觉的车流量统计方法来获取道路交通流参数，并使用优化的多目标智能配时优化算法求出对应不同流量比下的路口配时方案。本文研究工作主要分为三个部分：

       首先，在多车道车辆计数部分，为了满足智能交通系统对实时性和准确性的需求，通过横向对比其他五种不同算法，最终选择将YOLOv5作为基准网络用做车辆检测算法，融合改进ViBE背景建模算法的YOLOv5车辆检测方法，引入数据增强、Ghostnet、注意力机制，加强特征提取，提高模型的准确率与鲁棒性。利用人工标注虚拟线圈，根据合理的车道计数规则得到车辆的投影中心点来统计各车道车辆数目。实验表明，优化后的模型多车道计数准确率达到了97.7%，能够满足现实场景的需求。

       其次，在道路交叉口配时部分，首先介绍了多目标优化的原理，指出了固定配时方案依赖人工经验，同时建立了多目标配时优化算法模型；然后对经典NSGAⅡ算法进行优化，利用优化后的算法对配时数学模型进行求解，求出在不同流量比下优化后的配时方案；最后使用多个检验函数来验证改进NSGAⅡ多目标优化算法实时性、收敛性、分布度、综合性能评价指标。实验表明，优化后的算法相比原始算法FPS平均缩短了30.5%，△ 平均优化了23.8%，GD平均减少了52.3%，IGD提升了48.1%。

       最后，将求得的多目标配时优化方案设置VISSIM道路信息数据，与固定配时、Webster算法、NSGAⅡ和两种文献算法求出的配时方案在不同饱和度下进行比较，车均延误、车辆平均排队长度和通行能力三种指标表现优异。其中在流量比最大的高峰期，车均延误、车辆平均排队长度和通行能力得到了显著的提升，相较于固定配时算法，本文车均延误降低了27.5%、平均车辆排队长度降低了37.6%、通行能力提高了2.3%，能够有效缓解高峰时间段的交通拥堵问题。

With the development of the social economy, the number of motor vehicles in China is constantly increasing. The current fixed timing plan is difficult to match the complex road traffic conditions, and the problem of traffic congestion is becoming more and more serious. This has increased commuting time, raised the incidence of traffic accidents, and exacerbated environmental pollution. In this paper, a vehicle flow statistics system based on machine vision is proposed to obtain road traffic flow parameters, and an optimized multi-objective intelligent timing optimization algorithm is used to determine a reasonable intersection timing plan for different traffic flow ratios. This approach has become a hot research topic among scholars. The research work in this paper is mainly divided into three parts：

Firstly, in the multi-lane vehicle counting part, in order to meet the demand of ITS for real-time and accuracy, YOLOv5 is finally chosen as the benchmark network to be used as the vehicle detection algorithm by cross-sectional comparison of five other different algorithms, incorporating the YOLOv5 vehicle detection method with improved ViBE background modeling algorithm, introducing data enhancement, Ghostnet, and attention mechanism to enhance feature extraction to improve the accuracy and robustness of the model. Using manually labeled virtual coils, the projected centroids of vehicles are obtained to count the number of vehicles in each lane according to reasonable lane counting rules. Experiments show that the optimized model achieves an accuracy of 97.7% for multi-lane counting, which can meet the requirements of realistic scenarios.

Secondly, in the timing allocation section of road intersections, the principle of multi-objective optimization was introduced, pointing out the dependence of fixed timing plans on human experience, and a multi-objective adaptive timing optimization algorithm was constructed. Then, in view of the shortcomings of the classic NSGA II algorithm, it was optimized and used to solve the timing mathematical model, obtaining optimized timing plans under different traffic flow ratios. Finally, a classic verification function was used to test and improve the real-time performance, convergence, distribution, and overall performance evaluation indicators of the NSGA II multi-objective optimization algorithm. The experiments showed that the optimized algorithm reduced the average FPS by 30.5%, △optimized it by an average of 23.8%, GD  reduced it by an average of 52.3%, and IGD improved it by 48.1% compared to the original algorithm.

Finally, the obtained multi-objective timing optimization scheme is set with VISSIM road information data, and compared with the schemes proposed by fixed timing, Webster algorithm, and NSGA II under different saturation levels. The three indicators of vehicle average delay, vehicle average queue length, and traffic capacity perform well. In the peak period with the largest flow ratio, average vehicle delay, parking rate, and traffic capacity have been significantly improved. Compared to the fixed timing algorithm, the average vehicle delay in this article has been reduced by 27.5%, the queue length has been reduced by 37.6%, and the traffic capacity has been improved by 2.3%, which can effectively alleviate the traffic congestion problem during peak hours.

[1] 张阿嫱. 城镇化率达65.2%，城市迎新发展阶段[N]. 中国城市报,2023-03-06(003).

[2] 全国机动车保有量达4.17亿辆驾驶人超过5亿人[J].道路交通管理,2023,No.461(01):5.

[3] 朱波,张纪伟,谈东奎等.基于多源传感器与导航地图的端到端自动驾驶方法[J].汽车安全与节能学报,2022,13(04):738-749.

[4] 仇保兴.简论我国健康城镇化的几类底线[J].城市规划,2014,38(01):9-15.

[5] 孙世达, 金嘉欣, 吕建华, 等. 基于精细化年均行驶里程建立机动车排放清单[J]. 中国环境科学, 2020, 40(5): 2018-2029.

[6] Hadi, M. R., Kamboh, S. S., & Ali Shah, S. M. An intelligent transportation system for traffic safety and efficiency: a review of the literature. International Journal of Transportation Science and Technology, 2018,7(1), 1-15.

[7] 任寅.云顶之约：阿里云产业数字化“集智”之路[J].数字经济,2022,No.18(06):84-88.

[8] 李晓静. 自动驾驶汽车驾驶人侵权责任研究[D].山东大学,2022.

[9] 雷代文. 智能交通信号灯配时策略研究[D].重庆大学,2018.

[10] 李文举,姚建国,董天祯等.基于交错帧差和高斯模型的运动车辆检测[J].计算机工程与设计,2017,38(05):1347-1351.

[11] Qiming Z, ChengQian M. A vehicle detection method in tunnel video based on vibe algorithm[C]//2017 IEEE 2nd Advanced Information Technology, Electronic and Automation Control Conference (IAEAC). IEEE, 2017: 1545-1548.

[12] 张应辉, 刘养硕. 基于帧差法和背景差法的运动目标检测[J]. 计算机技术与发展, 2017, 27(2): 25-28.

[13] 陆子幸,宋健,王凯等.基于改进帧差法的视觉监控实验系统[J].实验室研究与探索,2018,37(09):21-24+28.

[14] 张彩珍,张云霞,陈永刚等.基于五帧差分和模板匹配的运动目标检测识别[J].激光与红外,2018,48(12):1554-1560.

[15] 段锁林,高仁洲,刘福等.融合改进帧差和视觉背景提取算法的前景检测[J].小型微型计算机系统,2019,40(09):1903-1908.

[16] Xia X, Lu X, Cao Y, et al. Moving vehicle detection with shadow elimination based on improved ViBe algorithm[C]//Journal of Physics: Conference Series. IOP Publishing, 2019, 1302(2): 022080.

[17] Sun W, Du H, Ma G, et al. Moving vehicle video detection combining ViBe and inter-frame difference[J]. International Journal of Embedded Systems, 2020, 12(3): 371-379.

[18] Xu D, Han G. Application of Improved ViBe Algorithm in Vehicle Detection[C]//2021 4th International Conference on Artificial Intelligence and Pattern Recognition. 2021: 199-204.

[19] Wieczorek G, Tahir S B U D, Akhter I, et al. Vehicle detection and recognition approach in multi-scale traffic monitoring system via graph-based data optimization[J]. Sensors, 2023, 23(3): 1731.

[20] Asvadi A, Garrote L, Premebida C, et al. Multimodal vehicle detection: fusing 3D-LIDAR and color camera data[J]. Pattern Recognition Letters, 2018, 115: 20-29.

[21] Rahman Z, Ami A M, Ullah M A. A real-time wrong-way vehicle detection based on YOLO and centroid tracking[C]//2020 IEEE Region 10 Symposium (TENSYMP). IEEE, 2020: 916-920.

[22] 丁冰,杨祖莨,丁洁等.基于改进YOLOv3的高速公路隧道内停车检测方法[J].计算机工程与应用,2021,57(23):234-239.

[23] Zhao M, Zhong Y, Sun D, et al. Accurate and efficient vehicle detection framework based on SSD algorithm[J]. IET Image Processing, 2021, 15(13): 3094-3104.

[24] 顾燕,李臻,杨锋等.基于改进Faster R-CNN的复杂背景红外车辆检测算法[J].激光与红外,2022,52(04):614-619.

[25] Singh C H, Mishra V, Jain K, et al. FRCNN-Based Reinforcement Learning for Real-Time Vehicle Detection, Tracking and Geolocation from UAS[J]. Drones, 2022, 6(12): 406.

[26] 陈家栋,雷斌.基于改进SSD轻量化的交通路口目标检测[J].传感器与微系统,2022,41(10):117-121+126.

[27] 潘卫华,罗裕坤.基于改进YOLOv3的车辆目标检测[J].计算机应用与软件,2023,40(01):167-172+204.

[28] Barbu T. Deep learning-based multiple moving vehicle detection and tracking using a nonlinear fourth-order reaction-diffusion based multi-scale video object analysis[J]. Discrete and Continuous Dynamical Systems-S, 2023, 16(1): 16-32.

[29] 袁建华,赵永进.我国道路交通信号控制的发展与变迁[J].道路交通管理,2022,No.451(03):10-12.

[30] 王逸,姚志洪,蒋阳升等.基于自适应遗传算法的双环信号配时优化模型[J].工业工程,2018,21(05):72-80.

[31] Guo J, Kong Y, Li Z, et al. A model and genetic algorithm for area-wide intersection signal optimization under user equilibrium traffic[J]. Mathematics and Computers in Simulation, 2019, 155: 92-104.

[32] 王逸,姚志洪,蒋阳升等.基于自适应遗传算法的双环信号配时优化模型[J].工业工程,2018,21(05):72-80.

[33] 邢岩,郝雨晴,高治军等.基于模糊控制的单交叉口信号配时优化[J].中国科技论文,2021,16(08):890-894.

[34] Mao T, Mihăită A S, Chen F, et al. Boosted genetic algorithm using machine learning for traffic control optimization[J]. IEEE Transactions on Intelligent Transportation Systems, 2021, 23(7): 7112-7141.

[35] 李娟,贾庭勇.基于人均延误的交叉口信号配时优化[J].公路交通科技,2021,38(11):134-141.

[36] Lin H, Han Y, Cai W, et al. Traffic signal optimization based on fuzzy control and differential evolution algorithm[J]. IEEE Transactions on Intelligent Transportation Systems, 2022.

[37] 任安虎,任洋洋,王瑶.多指标优化的深度强化学习单交叉口信号控制[J].国外电子测量技术,2022,41(10):104-111.

[38] 亢洁,李晓静.基于均值背景与三帧差分的运动目标检测[J].陕西科技大学学报,2018,36(01):148-153

[39] 韩雪莲,田爱奎,王振等.迭代自组织哈希算法[J].计算机应用研究,2020,37(05):1416-1420.

[40] 王立玲,刘超杰,马东等.改进自适应混合高斯模型和帧间差分的运动目标检测[J].机床与液压,2022,50(21):26-32.

[41] Kalirajan K, Sudha M. Moving object detection using median-based scale invariant local ternary pattern for video surveillance system[J]. Journal of Intelligent & Fuzzy Systems, 2017, 33(3): 1933-1943.

[42] Han K, Wang Y, Tian Q, et al. Ghostnet: More features from cheap operations[C]//Proceedings of the IEEE/CVF conference on computer vision and pattern recognition. 2020: 1580-1589.

[43] Qi J, Liu X, Liu K, et al. An improved YOLOv5 model based on visual attention mechanism: Application to recognition of tomato virus disease[J]. Computers and electronics in agriculture, 2022, 194: 106780.

[44] Du L, Lu Z, Li D. Broodstock breeding behaviour recognition based on Resnet50-LSTM with CBAM attention mechanism[J]. Computers and Electronics in Agriculture, 2022, 202: 107404.

[45] Xie F, Lin B, Liu Y. Research on the coordinate attention mechanism fuse in a YOLOv5 deep learning detector for the SAR ship detection task[J]. Sensors, 2022, 22(9): 3370.

[46] Gao J, Chen Y, Wei Y, et al. Detection of specific building in remote sensing images using a novel YOLO-S-CIOU model. Case: gas station identification[J]. Sensors, 2021, 21(4): 1375.

[47] Punn N S, Sonbhadra S K, Agarwal S, et al. Monitoring COVID-19 social distancing with person detection and tracking via fine-tuned YOLO v3 and Deepsort techniques[J]. arXiv preprint arXiv:2005.01385, 2020.

[48] 卢朝晖.基于视频虚拟检测线特征的交通流参数检测[J].电子技术与软件工程,2020,No.175(05):148-150.

[49] 尹聪慧,杨曦,彭强等.道路交叉口多目标信号配时优化设计[J].武汉理工大学学报,2022,44(08):84-90+96.

[50] 贾兴利,秦雪芳,周吴啸等.公路坡度差对车辆尾气排放的影响分析[J].科学技术与工程,2022,22(03):1265-1270.

[51] 郭运聪,韩晓龙.基于改进NSGA Ⅱ的时变路网污染路径问题多目标优化[J].上海海事大学学报,2022,43(04):60-66+90.

[52] 党根叶,苏宏升,车玉龙.基于改进NSGA-Ⅱ算法的碟式斯特林系统多目标优化研究[J].云南大学学报(自然科学版),2021,43(02):270-279.

[53] Katoch S, Chauhan S S, Kumar V. A review on genetic algorithm: past, present, and future[J]. Multimedia Tools and Applications, 2021, 80: 8091-8126.

[54] 徐炜铃. 县域城镇居民通勤交通需求分析研究[D].东南大学,2020.

[55] 王晓飞,丁振中,刘永等.基于Vissim建模的高速公路作业区混合车辆上游过渡区空间安全性研究[J].公路交通科技,2022,39(12):163-170.

[56] 曹海峰,叶朝辉.交通信号配时方案实时仿真系统设计与实现[J].传感器与微系统,2014,33(12):91-93+97.

[57] 何兆成,杨文臣,梁诗等.城市单交叉路口交通信号实时优化控制与仿真[J].计算机工程与应用,2010,46(33):239-243.