PotholeEye^(+): Deep-Learning Based Pavement Distress Detection System toward Smart Maintenance

We propose a mobile system,called PotholeEye+,for automatically monitoring the surface of a roadway and detecting the pavement distress in real-time through analysis of a video.PotholeEye+pre-processes the images,extracts features,and classifies the distress into a variety of types,while the road manager is driving.Every day for a year,we have tested PotholeEye+on real highway involving real settings,a camera,a mini computer,a GPS receiver,and so on.Consequently,PotholeEye+detected the pavement distress with accuracy of 92%,precision of 87%and recall 74%averagely during driving at an average speed of 110 km/h on a real highway.

Research on the Evaluation System of Epoxy Asphalt Steel Deck Pavement Distress Condition

Epoxy asphalt concrete has been one of the mainstream technology of steel deck pavement in China. But little specification about evaluation system for its distress condition has been researched and maintenance was still unsystematic. The section weight coefficient of different distress is proposed by analyzing the applicability of the “Highway Performance Assessment Standards”. Indexes mainly including SDPCI PDR and PCR are presented to evaluate its distress condition. The evaluation system and maintenance plan decision tree were recommended which can assist scientific maintenance of epoxy asphalt steel deck pavement.

Automated classification and detection of multiple pavement distress images based on deep learning

To achieve automatic,fast,efficient and high-precision pavement distress classification and detection,road surface distress image classification and detection models based on deep learning are trained.First,a pavement distress image dataset is built,including 9017pictures with distress,and 9620 pictures without distress.These pictures were captured from 4 asphalt highways of 3 provinces in China.In each pavement distress image,there exists one or more types of distress,including alligator crack,longitudinal crack,block crack,transverse crack,pothole and patch.The distresses are labeled by a rectangle bounding box on the pictures.Then ResNet networks and VGG networks are used respectively as binary classification models for distressed and non-distressed imagines classification,and as multi-label classification models for six types of distress classification.Training techniques,such as data augmentation,batch normalization,dropout,momentum,weight decay,transfer learning,and discriminative learning rate are used in training the model.Among the 4 CNNs considered in this study,namely ResNet 34 and 50,and VGG 16 and 19,for the binary classification,ResNet 50 has the highest Accuracy of 96.243%,Precision of 95.183%,and ResNet 34 has the highest Recall of 97.824%,and F2 score of 97.052%.For multi-label classification,ResNet 50 has the best performance,with the highest Accuracy of 90.257%,higher than 90%required by the Chinese standard(JTG H20-2018)for road distresses detection,F2 score-82.231%,and Precision-76.509%,and ResNet34 has the highest Recall of 87.32%.To locate and quantify the distress areas in the images,the single shot multibox detector(SSD)model is developed,in which the ResNet 50 is used as the base network to extract features.When the intersection over union(IoU)is set to 0,0.25,0.50,0.75,the mean average precision(mAP)of the model are found to be 74.881%,50.511%,28.432%,3.969%,respectively.

A review on pavement distress and structural defects detection and quantification technologies using imaging approaches

Pavement distress detection(PDD)plays a vital role in planning timely pavement maintenance that improves pavement service life.In order to promote the development of PDD technologies and find out the insufficiencies in PDD field,this paper reviews the technical development history and characteristics of various PDD technologies,which contributes to the current state of research on PDD.First,processes of PDD are briefly introduced.The PDD technologies based on radar ranging,2D image,laser ranging and 3 D structured light are illustrated.The newest 3D PDD technology based on interference fringe,which has better accuracy,is in progress.The principles and implementation processes of these methods are discussed.Finally,the shortcomings of these technologies in the field of PDD are concluded.Recommendations for future development are provided.The research results show that various PDD technologies have been continuously improved,developed,over the past decade,and have achieved a series of results.However,the measurements from existing PDD technologies can not be metrological traced to acquire the true dimensions of pavement distresses.The lack of metrological traceability technology in the PDD field needs to be further solved.In order to achieve more accurate and efficient PDD,the metrological traceability technology of PDD systems has become the future development direction in this field.

Review:Asphalt Pavement Rutting Distress and Affects on Traffics Safety

Highway is an essential facility that led to both economic success and quality of life.Maintenance is necessary to ensure that highway will able to continue to carry out its functions.The rutting of asphalt pavement structures during their exploitation is considered to be one of the main problems in the entire world.This kind of pavement distress makes a negative impact to the exploitation characteristics of the asphalt pavement to the residual life of pavement structure,also to the safety and quality of the traffic.The main purpose of this review is to define the effects of rutting on roads safety.

Predicting pavement condition index based on the utilization of machine learning techniques:A case study

Pavement management systems(PMS)are used by transportation government agencies to promote sustainable development and to keep road pavement conditions above the minimum performance levels at a reasonable cost.To accomplish this objective,the pavement condition is monitored to predict deterioration and determine the need for maintenance or rehabilitation at the appropriate time.The pavement condition index(PCI)is a commonly usedmetric to evaluate the pavement's performance.This research aims to create and evaluate prediction models for PCI values using multiple linear regression(MLR),artificial neural networks(ANN),and fuzzy logic inference(FIS)models for flexible pavement sections.The authors collected field data spans for 2018 and 2021.Eight pavement distress factors were considered inputs for predicting PCI values,such as rutting,fatigue cracking,block cracking,longitudinal cracking,transverse cracking,patching,potholes,and delamination.This study evaluates the performance of the three techniques based on the coefficient of determination,root mean squared error(RMSE),and mean absolute error(MAE).The results show that the R2 values of the ANN models increased by 51.32%,2.02%,36.55%,and 3.02%compared toMLR and FIS(2018 and 2021).The error in the PCI values predicted by the ANNmodel was significantly lower than the errors in the prediction by the FIS and MLR models.

Study on Ground-Penetrating Radar (GPR) Application in Pavement Deep Distress Detection

As some deep distresses exist in pavement structures, ground-penetrating radar (GPR) reflected waves will vary at interfaces and defects. Aimed at detecting the distresses in terms of position, severity and degree, electromagnetic forward simulations based on 400 MHz and 900 MHz antennas were conducted respectively. The dielectric models concerning homogeneous or coupling distresses of pavements were established, and the effects of various distresses on detection were analyzed through reflected wave images. Relying on GPR tests and field tests, coring and excavation data acquired before rehabilitation were compared and verified. The calculation results match the field measurement results. Thus, the detection method based on GPR was proposed for pavement deep distresses.

Damage Mechanism of Ultra-thin Asphalt Overlay(UTAO) based on Discrete Element Method

Aiming to analyze the damage mechanism of UTAO from the perspective of meso-mechanical mechanism using discrete element method(DEM),we conducted study of diseases problems of UTAO in several provinces in China,and found that aggregate spalling was one of the main disease types of UTAO.A discrete element model of UTAO pavement structure was constructed to explore the meso-mechanical mechanism of UTAO damage under the influence of layer thickness,gradation,and bonding modulus.The experimental results show that,as the thickness of UTAO decreasing,the maximum value and the mean value of the contact force between all aggregate particles gradually increase,which leads to aggregates more prone to spalling.Compared with OGFC-5 UTAO,AC-5 UTAO presents smaller maximum and average values of all contact forces,and the loading pressure in AC-5 UTAO is fully diffused in the lateral direction.In addition,the increment of pavement modulus strengthens the overall force of aggregate particles inside UTAO,resulting in aggregate particles peeling off more easily.The increase of bonding modulus changes the position where the maximum value of the tangential force appears,whereas has no effect on the normal force.

移动荷载下道路病害的弯沉盆特性

移动荷载作用下道路弯沉盆蕴含路基、路面的病害信息,充分利用移动弯沉盆信息开展道路病害快速无损检测受到行业重要关注。该文通过分析道路病害的弯沉盆特性,为移动弯沉车开发道路病害检测方法提供技术参考。首先建立半刚性基层沥青路面有限元模型,通过设置不同类型和程度的病害有限元模型,分析在移动荷载作用下路表弯沉盆的连续变化情况。结果表明:荷载移动到病害区域时,各病害模型中路表弯沉盆的中心弯沉值均会变大;路面结构层出现横向裂缝时,弯沉盆的形状会发生明显变化,贯通性横向裂缝对弯沉盆形状的影响最明显;基层与底基层局部松散对弯沉盆形状的影响比面层松散对弯沉盆的影响更明显;土基顶部出现空洞时,弯沉盆形状变化不明显;路基唧泥病害会严重影响路表弯沉盆形状与大小。

内容感知的可解释性路面病害检测模型

针对实际场景中高分辨路面图像难以直接作为现有卷积神经网络(convolutional neural network,CNN)的输入、现有预处理及下采样算法无法有效感知并保留原始路面图像中低占比的病害区域信息等问题,借助于可视化解释的技术手段,设计了一种即插即用的图像内容自适应感知模块(adaptive perception module,APM),既平衡了高分辨路面图像与CNN输入限制,又能够自适应感知激活前景病害区域,从而实现高分辨路面图像中病害类型的快速准确检测,构建可信路面病害视觉检测软件系统.APM利用大卷积核和下采样残差操作降低原始图像分辨率并获取图像浅层特征表示;通过注意力机制自适应感知并激活图像中路面病害区域信息,过滤无关的背景信息.利用联合学习的方式,无需额外监督信息完成对APM的训练.通过可视化解释方法辅助选择和设计APM的具体结构,在最新公开数据集CQUBPMDD上的实验结果表明:APM相比于现有的图像预处理采样算法均有明显提升,分类准确率最高为84.47%;在CQU-BPDD上的实验结果及APM决策效果可视化分析表明APM具备良好的泛化性与鲁棒性.实验代码已开源:https://github.com/Li-Ao-Git/apm.

基于改进YOLOv7的路面病害检测算法

针对公路路面病害图像存在光影变化大、背景干扰多、尺度差异大等问题,提出基于改进YOLOv7的路面病害检测算法。首先,对YOLOv7网络模型中的ELAN模块进行了优化,通过通道和空间注意力机制优化信息提取,增强网络对重要特征的提取能力;接着,使用ACmix注意力模块提高网络对小目标的关注度,有效解决原网络模型对小目标的漏检问题;其次,采用大下采样比率的卷积输出,提高对小目标的检测精度;最后,引入WIoUv3替换原网络模型中的CIoU来优化损失函数,构造梯度增益的计算方法来附加聚焦机制。实验结果表明:改进后的YOLOv7模型平均精度均值(mAP)与原模型相比提升了4.5%,检测效果优于原网络模型与传统经典目标检测网络模型。

Structural Design of Pavement Overlays Based on Functional Parameters

This paper reports a practical pavement overlay design method based on PCI (Pavement Condition Index). Current pavement investigation method (JTJ 073 96) is compared to the ASTM D 5340, which is the standard test method for airport pavement condition evaluation initially developed for US Air Force. The deficiency in the calculation of PCI based on field data in JTJ 073 is discussed. The proposed design method is compared to AASHTO overlay design method with good agreement. The paper concludes with an example illustrating how the existing pavement structural capacity is related to pavement distress survey results. The presented design method can be used in the design for overlay rehabilitation of pavements of highways, urban streets and airports.

Image Preprocessing Methods to Identify Micro-cracks of Road Pavement

Standards of highway conservation and maintenance are improved gradually following the improvement of requirements of road service. Before obvious damage such as obvious cracking (block,transverse, longitudinal ) and rutting emerge, inconspicuous distress (micro-cracks, polishing, pockmarked) is generated previously. These inconspicuous distresses may provide basis and criteria for pavement preventive maintenance. Currently most of preventive conservation measures are determined by experienced experts in maintenance and repair of road after site visits. Thus method is difficult in operation, and has a certain amount of instability as it is based on experience and personal knowledge. In this paper, camera and laser were used for automated high-speed acquisition images. Methods to preprocess pavement image are compared. The pretreatment method suitable for analyze micro-cracks picture is elected, an effective way to remove shadow is also proposed.

基于改进YOLOv5-DeepSORT算法的公路路面病害智能识别

针对目标检测算法在多尺度无人机(UAV)图像中对公路路面病害容易出现漏检误检及同一病害在连续帧图片中被重复检测的问题,提出一种具有病害重识别能力的路面检测方法。通过引入真实宽高损失与纵横比以提升损失函数性能,利用CA(Coordinate attention)注意力机制提升模型在复杂背景下的识别能力,将模型初始特征引入特征融合网络提升模型检测多尺度病害的鲁棒性,构建基于DeepSORT(目标检测的多目标跟踪算法)的二级检测机制实现对病害的重识别与统计。试验结果表明,模型平均检测精度mAP达到89.19%,较基准模型提升了3.11%;F1分数为0.8514,较原模型提升了2.49%;同时也优于主流目标检测算法;在无人机影像下病害计数精度达到91.38%,较改进前提升25.86%,为公路路面检测与养护提供精确实时的病害数据。

基于多分支深度学习的沥青路面多病害检测方法

为了准确、快速地识别路面多病害,采用一种基于多分支框架的深度学习方法,提取并融合路面图像的大、小尺度特征,将路面二维图像和三维图像作为网络输入,增强病害特征.采集裂缝、条状修补、块状修补、坑槽、松散等沥青路面病害图像共计10 562张,进行人工标注.结果表明:500次训练后该方法的平均交并比为0.83,准确率和召回率的调和平均数F值为0.90,优于U-net、PSPNet、DeepLabv3+等方法;在单一类别上,对条状修补、坑槽、松散、桥接缝等分割效果最优,对裂缝、块状修补的识别展现出较强的鲁棒性;所提方法的识别效果高于仅使用单一输入或者单一分支的方法.因此,双通道和多分支的设计方法可以显著提升网络对多类别路面病害的识别精度.

基于改进SSD模型的路面病害识别算法研究

为解决公路路面病害图像特征不突出、检测精度偏低的问题,提出了一种改进SSD模型的路面病害识别算法.在SSD网络结构的基础上将其基础网络替换为Dense-net网络,使得特征信息更加容易被获取,并能够降低网络参数数量.同时在算法中增添了注意力机制,加强有用特征的利用效率.为了更好地观察算法改进的效果,不仅在已知的路面数据集上进行了测试,还在自制的数据集上进行了测试.从测试结果来看,SSD模型改进后在两种数据集上的分类准确度分别为93.5%和90.28%,比原SSD300模型分别提高了4.8%和6.36%,说明该模型能够有效的提升病害识别的准确性.

公路水泥混凝土路面破坏机理及防治措施

水泥混凝土路面在实际使用中受诸多因素的作用,产生多种病害,影响公路服务质量,威胁行车安全。论文总结水泥混凝土路面常见的病害类别和产生机理,并以重庆地区某水泥混凝土路面工程为例,提出针对各病害的治理与预防方案,从而延长水泥混凝土路面的服务年限。

基于YOLOX-Resnet50模型的路面病害识别方法的应用与验证

为降低路面病害检测的人工和时间成本,提高检测效率,文章基于采集的G320国道某路段的病害数据,运用改进的YOLOX-Resnet50模型对路面病害进行智能识别,通过两轮训练(其中第二轮是在第一轮数据集的基础上优选数据集),以评估该模型的性能,探究YOLOX-Resnet50模型在路面病害检测方面的适用性。研究表明:YOLOX-Resnet50模型在路面检测中对病害检测准确度分别为车辙75%、标线67%、修补67%、纵向裂缝50%、横向裂缝40%、网裂24%,平均准确度达到了53.8%。该模型能初步替代人工对病害进行筛选,可有效减少人工及时间成本。

路面破损自动识别的一种新算法

针对路面破损分类这一难题,提出了一种基于破损密度因子的路面破损分类新算法。对破损密度因子进行了定义和仿真实验,仿真结果表明其对5种常见的路面破损状况的分类效果非常理想。为了进行对比,文中还选择了美国博士论文中的PROXIMITY算法进行比较,两种方法对相同的10几万幅路面样本进行分类试验,试验结果表明,笔者提出的基于破损密度因子的路面破分类方法,整体优于PROXIMITY方法的分类效果。

基于不变矩特征的沥青路面破损图像识别

提出了一种减少沥青路面破损图像识别计算量的图像分割方法。将路面图像等分为64×64像素的子块图像,并用灰度方差值描述子块图像特征。设计了基于BP神经网络的子块图像模式分类器,利用子块图像模式分类结果所组成的矩阵作为路面破损图像分割结果。并将路面破损图像子块模式矩阵的不变矩作为路面破损图像的整体特征,在此基础上设计了基于全局优化算法的路面破损前馈神经网络分类器。最后进行了路面破损图像识别试验,识别率达到83 3%。