{"title":"使用机器学习的自适应交通灯系统","authors":"M. Ottom, A. Al-Omari","doi":"10.34028/iajit/20/3/13","DOIUrl":null,"url":null,"abstract":"Traffic congestion is a major problem in many cities of the Hashemite Kingdom of Jordan as in most countries. The rapidly increase of vehicles and dealing with the fixed infrastructure have caused traffic congestion. One of the main problems is that the current infrastructure cannot be expanded further. Therefore, there is a need to make the system work differently with more sophistication to manage the traffic better, rather than creating a new infrastructure. In this research, a new adaptive traffic lights system is proposed to determine vehicles type, calculate the number of vehicles in a traffic junction using patterns detection methods, and suggest the necessary time for each side of the traffic junction using machine learning tools. In this context, the contributions of this paper are: (a) creating a new image-based dataset for vehicles, (b) proposing a new time management formula for traffic lights, and (c) providing literature of many studies that contributed to the development of the traffic lights system in the past decade. For training the vehicle detector, we have created an image-based dataset related to our work and contains images for traffic. We utilized Region-Based Convolutional Neural Networks (R-CNN), Fast Region-Based Convolutional Neural Networks (Fast R-CNN), Faster Region-Based Convolutional Neural Networks (Faster R-CNN), Single Shot Detector (SSD), and You Only Look Once v4 (YOLO v4) deep learning algorithms to train the model and obtain the suggested mathematical formula to the required process and give the appropriate timeslot for every junction. For evaluation, we used the mean Average Precision (mAP) metric. The obtained results were as follows: 78.2%, 71%, 75.2%, 79.8%, and 86.4% for SSD, R-CNN, Fast R-CNN, Faster R-CNN, and YOLO v4, respectively. Based on our experimental results, it is found that YOLO v4 achieved the highest mAP of the identification of vehicles with (86.4%) mAP. For time division (the junctions timeslot), we proposed a formula that reduces about 10% of the waiting time for vehicles.","PeriodicalId":13624,"journal":{"name":"Int. Arab J. Inf. Technol.","volume":"358 1","pages":"407-418"},"PeriodicalIF":0.0000,"publicationDate":"2023-01-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":"2","resultStr":"{\"title\":\"An adaptive traffic lights system using machine learning\",\"authors\":\"M. Ottom, A. Al-Omari\",\"doi\":\"10.34028/iajit/20/3/13\",\"DOIUrl\":null,\"url\":null,\"abstract\":\"Traffic congestion is a major problem in many cities of the Hashemite Kingdom of Jordan as in most countries. The rapidly increase of vehicles and dealing with the fixed infrastructure have caused traffic congestion. One of the main problems is that the current infrastructure cannot be expanded further. Therefore, there is a need to make the system work differently with more sophistication to manage the traffic better, rather than creating a new infrastructure. In this research, a new adaptive traffic lights system is proposed to determine vehicles type, calculate the number of vehicles in a traffic junction using patterns detection methods, and suggest the necessary time for each side of the traffic junction using machine learning tools. In this context, the contributions of this paper are: (a) creating a new image-based dataset for vehicles, (b) proposing a new time management formula for traffic lights, and (c) providing literature of many studies that contributed to the development of the traffic lights system in the past decade. For training the vehicle detector, we have created an image-based dataset related to our work and contains images for traffic. We utilized Region-Based Convolutional Neural Networks (R-CNN), Fast Region-Based Convolutional Neural Networks (Fast R-CNN), Faster Region-Based Convolutional Neural Networks (Faster R-CNN), Single Shot Detector (SSD), and You Only Look Once v4 (YOLO v4) deep learning algorithms to train the model and obtain the suggested mathematical formula to the required process and give the appropriate timeslot for every junction. For evaluation, we used the mean Average Precision (mAP) metric. The obtained results were as follows: 78.2%, 71%, 75.2%, 79.8%, and 86.4% for SSD, R-CNN, Fast R-CNN, Faster R-CNN, and YOLO v4, respectively. Based on our experimental results, it is found that YOLO v4 achieved the highest mAP of the identification of vehicles with (86.4%) mAP. For time division (the junctions timeslot), we proposed a formula that reduces about 10% of the waiting time for vehicles.\",\"PeriodicalId\":13624,\"journal\":{\"name\":\"Int. Arab J. Inf. Technol.\",\"volume\":\"358 1\",\"pages\":\"407-418\"},\"PeriodicalIF\":0.0000,\"publicationDate\":\"2023-01-01\",\"publicationTypes\":\"Journal Article\",\"fieldsOfStudy\":null,\"isOpenAccess\":false,\"openAccessPdf\":\"\",\"citationCount\":\"2\",\"resultStr\":null,\"platform\":\"Semanticscholar\",\"paperid\":null,\"PeriodicalName\":\"Int. Arab J. Inf. Technol.\",\"FirstCategoryId\":\"1085\",\"ListUrlMain\":\"https://doi.org/10.34028/iajit/20/3/13\",\"RegionNum\":0,\"RegionCategory\":null,\"ArticlePicture\":[],\"TitleCN\":null,\"AbstractTextCN\":null,\"PMCID\":null,\"EPubDate\":\"\",\"PubModel\":\"\",\"JCR\":\"\",\"JCRName\":\"\",\"Score\":null,\"Total\":0}","platform":"Semanticscholar","paperid":null,"PeriodicalName":"Int. Arab J. Inf. Technol.","FirstCategoryId":"1085","ListUrlMain":"https://doi.org/10.34028/iajit/20/3/13","RegionNum":0,"RegionCategory":null,"ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":null,"EPubDate":"","PubModel":"","JCR":"","JCRName":"","Score":null,"Total":0}
引用次数: 2
摘要
与大多数国家一样,交通堵塞是约旦哈希姆王国许多城市的一个主要问题。车辆的快速增加和对固定基础设施的处理造成了交通拥堵。其中一个主要问题是,目前的基础设施无法进一步扩大。因此,有必要让系统以不同的方式工作,更复杂地管理交通,而不是创建一个新的基础设施。在本研究中,提出了一种新的自适应交通灯系统来确定车辆类型,使用模式检测方法计算交通路口的车辆数量,并使用机器学习工具建议交通路口每侧的必要时间。在此背景下,本文的贡献是:(a)为车辆创建了一个新的基于图像的数据集,(b)提出了一个新的交通信号灯时间管理公式,以及(c)提供了过去十年中对交通信号灯系统发展做出贡献的许多研究的文献。为了训练车辆检测器,我们创建了一个与我们的工作相关的基于图像的数据集,其中包含交通图像。我们利用基于区域的卷积神经网络(R-CNN)、快速区域卷积神经网络(Fast R-CNN)、更快区域卷积神经网络(Faster R-CNN)、单镜头检测器(SSD)和You Only Look Once v4 (YOLO v4)深度学习算法对模型进行训练,得到所需过程的建议数学公式,并给出每个连接点的适当时间段。为了进行评估,我们使用了平均精度(mAP)度量。结果表明:SSD、R-CNN、Fast R-CNN、Faster R-CNN、YOLO v4分别为78.2%、71%、75.2%、79.8%、86.4%。根据我们的实验结果,YOLO v4的车辆识别mAP最高,为86.4%。对于时间划分(路口时隙),我们提出了一个公式,可以减少约10%的车辆等待时间。
An adaptive traffic lights system using machine learning
Traffic congestion is a major problem in many cities of the Hashemite Kingdom of Jordan as in most countries. The rapidly increase of vehicles and dealing with the fixed infrastructure have caused traffic congestion. One of the main problems is that the current infrastructure cannot be expanded further. Therefore, there is a need to make the system work differently with more sophistication to manage the traffic better, rather than creating a new infrastructure. In this research, a new adaptive traffic lights system is proposed to determine vehicles type, calculate the number of vehicles in a traffic junction using patterns detection methods, and suggest the necessary time for each side of the traffic junction using machine learning tools. In this context, the contributions of this paper are: (a) creating a new image-based dataset for vehicles, (b) proposing a new time management formula for traffic lights, and (c) providing literature of many studies that contributed to the development of the traffic lights system in the past decade. For training the vehicle detector, we have created an image-based dataset related to our work and contains images for traffic. We utilized Region-Based Convolutional Neural Networks (R-CNN), Fast Region-Based Convolutional Neural Networks (Fast R-CNN), Faster Region-Based Convolutional Neural Networks (Faster R-CNN), Single Shot Detector (SSD), and You Only Look Once v4 (YOLO v4) deep learning algorithms to train the model and obtain the suggested mathematical formula to the required process and give the appropriate timeslot for every junction. For evaluation, we used the mean Average Precision (mAP) metric. The obtained results were as follows: 78.2%, 71%, 75.2%, 79.8%, and 86.4% for SSD, R-CNN, Fast R-CNN, Faster R-CNN, and YOLO v4, respectively. Based on our experimental results, it is found that YOLO v4 achieved the highest mAP of the identification of vehicles with (86.4%) mAP. For time division (the junctions timeslot), we proposed a formula that reduces about 10% of the waiting time for vehicles.