Roads are the foundation of intelligent transportation systems, yet cracks are widely present in roads and seriously affect system performance. Cracks not repaired promptly can develop into severe road defects, significantly increasing the risk of traffic accidents. Researchers in the community have started to focus on the automatic sensing of cracks in asphalt pavements, while it is still a challenging task on concrete pavements. Cracks in the concrete pavement are easily recognized as interrupted segments rather than a continuous whole due to the interference of the surface texture. This mistake can seriously mislead the judgment of cracks and subsequent road repair. In this paper, we aim to solve the challenge by enhancing contextual information about cracks within the images. We first extract the information from the local and global representations using image information and then fuse it into complete contextual information by a designed multilayer perceptron. Finally, we use the discriminative loss to constrain the edges of cracks and backgrounds using complete crack contextual information. We have collected and annotated several images of concrete pavements from several significant provinces in China. Experiments show that our method achieves the best performance compared to state-of-the-art methods, especially in edge determination.

Automatic detection and segmentation of potholes and raveling are crucial for preventive maintenance and ensuring roads structural health. However, the extraction of pavement potholes was mainly based on traditional image processing methods, which have proven to be ineffective and inaccurate. Additionally, the absence of a unified pothole and raveling dataset has resulted in the lack of benchmarks for evaluating various methods. This work makes three contributions to address these problems. Firstly, we have curated the pavement pothole and raveling distress detection and segmentation datasets. Secondly, a two-step pavement pothole and raveling detection and segmentation method was proposed. In the initial step, an automated pavement pothole and raveling detection model was developed using the modified YOLOX. Subsequently, the segmentation model, named dual self-attention segmentation network (DSASNet), was proposed to segment distress by extracting mode-sensitive features from intensity and range images using two parallel Twins-SVT self-attention branches. Moreover, we design a mid-fusion module to adaptively fuse mode-specific and scale-specific features. Finally, a pyramid pooling module (PPM) is connected to further enhance the segmentation capability for potholes and raveling of various sizes and shapes. The F1-score and Intersection over union (IoU) of the proposed DSASNet on the test set are 93.65% and 0.881, respectively, outperforming other baseline methods. Furthermore, we conduct an experiment to quantitatively compare the two-step method with the one-step method using only a single semantic segmentation model. The results demonstrated clear advantages of the proposed two-step method in terms of accuracy and efficiency for pavement pothole and raveling segmentation.

Automated pavement crack detection is crucial to supporting fine pavement maintenance and ensuring safety for road facilities. Due to the complex pavement condition and crack features, it is still a critical challenge in intelligent pavement surveys. This paper proposed a novel pixel-level pavement crack segmentation network, PCSNet, to provide a solution to this challenge. The network has richer attention and hybrid pyramid structures, which implement full-process crack feature fusion and enhancement. The richer attention module consists of cascaded self-attention and attention gate modules. It captures the crack spatial dependence information and prunes the feature response. The hybrid pyramid structures consist of a multistage convolutional pyramid module and a pyramid pooling module. It integrates contextual information at multiple receptive field scales to enhance the potential crack feature representation. The proposed structure enriches the crack details and optimises the scene parsing on the global geometry of the cracks. A sizeable 3D pavement crack dataset is built for training and testing. The proposed network exhibited the best performance, achieving F1-score, mean intersection of union, and mean pixel accuracy of 81.21%, 77.13%, and 87.17%, respectively. The network can reconstruct the complete crack geometry, preserve the crack edges well, and optimises the detection of shallow and complex cracks. The method exhibits superior and robust performance, facilitating accurate pavement technical condition assessment and maintenance decisions.

Recently, many deep learning methods have achieved great results in the field of automated pavement distress detection, but most of them ignore other types of distresses beyond cracks. This paper proposes an efficient deep learning framework for automated asphalt pavement distress segmentation called pavement distress segmentation network (PDSNet). PDSNet can effectively segment multiple asphalt pavement distresses, including crack, pothole, raveling, patch, and sealed crack. It consists of two parallel feature extraction branches. One is the P branch to extract prior global information. The other is the U branch to obtain local information. By utilising the global and local features together, PDSNet can produce precise segmentation results under complicated circumstances. For deep learning purposes, a pavement distress dataset consisting of 4000 pavement images is collected and manually labelled at pixel level. Each image of the pavement dataset is a two-channel image, which is concatenated by a 2D pavement image and the correspondingly 3D pavement image. Particularly, it is the first pavement distress dataset that utilises two-channel pavement images. According to the experimental results, PDSNet yields a performance with a MIoU of 83.7%. Compared with the state-of-the-art networks, PDSNet achieves the best MIoU and has considerable parameter number and inference time.

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.

The detection of pavement crack plays a critical role in pavement maintenance and rehabilitation because pavement cracking is one of the most important indicators for the pavement condition evaluation, as well as an early manifestation of other pavement distresses. To detect cracks accurately, precisely, and completely based on three-dimensional (3D) pavement images, this paper proposes a deep learning framework based on a convolutional neural network (CNN) and pixel-level improved crack seed algorithm, called Pavement Crack Detection Net (PCDNet). Firstly, the CNN layer based on the convolution implementation of sliding windows is applied to each 3D pavement image to divide it into 8 x 8 pavement patches and classify each patch into two types: the background patch, and the pavement crack patch. Secondly, the seed layer, i.e., an automatic threshold pixel-level crack seed recognition algorithm is used to detect the crack distress further and depict the complete contour simultaneously. Finally, the region growing layer is utilized to ensure the continuity of the cracks. Due to the good combination of the CNN and the algorithm, PCDNet needs only a patch-level data set for training but can output pixel-level results, a great novelty in crack detection. In this paper, 5,000 3D pavement images were selected from an established image library. PCDNet was trained with 4,300 3D pavement images and further validated based on 500 3D pavement images. The test experiment based on the remaining 200 images showed that PCDNet can achieve high precision (0.885), recall (0.902), and F-1 score (0.893) simultaneously. It also was demonstrated that PCDNet can detect different types of pavement crack under various conditions and resist noncrack pixels with elevation variation features, such as pavement edge drop-offs, curbs, spalling, and bridge expansion joints. Compared with recently developed crack detection methods based on imaging algorithms, PCDNet is capable of not only eliminating more local noise and detecting more tine cracks, but also maintaining much faster processing speed. (C) 2022 American Society of Civil Engineers.

Revealing urban community structures of a city is of great importance for investigating urban development and sprawl behind the movement dynamics. However, most studies focus on delineating urban community structure and its variations with a single transit mode without covering hierarchical travel distances. This paper proposes an overarching framework to reveal urban community structures by fusing multisource spatiotemporal transportation data. Network science methods and community detection are applied to construct spatially embedded networks and uncover the urban structure from different perspectives, using 1-week transportation data derived from the metro, taxi, and dockless bike-sharing systems (BSSs) of Shanghai, China, in year 2016. Our finding shows that Shanghai can be clustered into six primary communities and exhibits polycentric patterns with strong monocentric characteristics. Shanghai's urban structure moves toward an embedded hierarchical pattern: the dispersed monocentric structure and the centralized polycentric structure. It reflects poor functional interdependence and horizontal connectivity between communities. Beneath the complex and coupled travel-flow system, the metro dominants the basic framework of the urban community structure and contributes to form the prototype of the core community, while the taxi and BSS tend to play complementary roles like expanding, enhancing, and refining the structure. This research not only provides a promising bridge from the complex urban transportation networks to urban community structures, but also implies potential urban planning policies from an internal and comprehensive perspective.

The construction of highways has been well-developed worldwide. Meanwhile, the heavy traffic flow brings huge pressure on highway maintenance. Pavement rutting is one of the major pavement distresses and its detection has been a research hot spot in pavement engineering. Despite the fruitful research outcomes, most of them were based on ideal circumstances and focused on how to improve the processing procedure to reduce the detection error of usual rutting measurement. Whereas some particular interference, such as pavement markings under strong light, usually occurs during the detection, and remains undetected. Pavement markings affect the accurate extraction of pavement transverse profiles and increase the detection error of rut depth. To fill this gap, this study proposed a line-structured rut detection method to improve the detecting accuracy of rut depth. The global gray scale correction algorithm and feature-based fusion segmentation algorithm are mainly used to eliminate pavement markings of the background. The centerline-based midpoint thinning algorithm, least square based curve correction method, and envelope model are applied to calculate the rut depth, and are applicable for different forms of rutting distress. A total of 600 of images collected from urban roads were classified into four categories and used to verify the proposed rut detection method with pavement markings interference under strong light. The experimental results indicate that the average relative detection error is 10.07% and the average proportion of detection accuracy is 87.65%. Meanwhile, the evaluation accuracy of the pavement condition assessed by the rut depth index reaches 83.87%. This manifests that the proposed method can not only deal with the rutting detection with interference, but can also apply to the situation without interference. Thus, the method could be used to evaluate pavement condition and offer a reliable data source for pavement maintenance. The work in the paper offers a vital reference for pavement rut detection methods worldwide.

Crack is a common concrete pavement distress that will deteriorate into severe problems without timely repair, which means the automated detection of pavement crack is essential for pavement maintenance. However, automatic crack detection and segmentation remain challenging due to the complex pavement condition. Recent research on pavement crack detection based on deep learning has laid a good foundation for automated crack segmentation, but there can still be improvements. This paper proposes an automatic concrete pavement crack segmentation framework with enhanced graph network branch. First, the nodes of the graph and nodes’ attributions are generated based on the image dividing. The edges of the graph are determined based on Gaussian distribution. Then, the graph from the image is input into the graph branch. The graph feature map of the graph branch output is fused with the image feature map of the encoder and then enters the decoder to recover the image resolution to obtain the crack segmentation result. Finally, the method is tested on a self-built 3D concrete pavement crack dataset. The proposed method achieves the highest F1 and IoU (Intersection over Union) in the comparison experiments. And the graph branch addition improves 0.08 on F1 and 0.06 on IoU compared with U-Net.

为了在路面三维图像的基础上快速、准确、完整地识别裂缝,提出一种基于深度学习的路面裂缝类病害自动检测方法.首先,以子块图像为处理单元,将三维图像划分为裂缝面元和背景面元,其中背景面元包含了路面标线、不同纹理和桥接缝等复杂场景.根据对面元图像的分析,提出一种基于卷积神经网络的PCCNet分类模型,用于路面背景面元和路面裂缝面元的自动识别.然后,为了进一步提取裂缝面元内裂缝的完整轮廓,考虑路面三维图像中裂缝像素级邻域特征,利用PCCNet模型结合裂缝高程检查方法对路面裂缝进行检测.研究结果表明:通过训练集4 300张高精度三维图像的训练,模型在3 850次迭代之后出现过拟合,且此时PCCNet模型在验证集上的总体F值达到最大,为92.9％;将PCCNet模型结合裂缝高程检查方法应用在测试集的200张三维图像上,方法准确率、召回率和F值分别为87.8％、90.1％和88.9％.与改进Canny方法和种子识别方法对比,所提出的方法在抑制噪声和检测细小裂纹方面具有更强的鲁棒性.

目的 在城市交通检测中,智能视频的广泛应用使得人工智能技术及计算机视觉先进技术对视频中的前景目标检索、识别、特征提取、行为分析等成为视觉研究的热点,但由于复杂场景中动态背景具有不连续的特点,使得少部分的前景目标图像信息丢失,从而造成漏检、误判.方法 本文提出一种RPCA(鲁棒主成分分析)优化方法,为了快速筛选与追踪前景目标,以基于帧差欧氏距离方法设计显著性目标帧号快速提取算法,确定关键帧邻域内为检测范围,对经过稀疏低秩模型初筛选的前景目标图像进行前景目标种子并行识别和优化连接,去除前景目标图像中的动态背景,同时将MASK(掩膜)图像中的前景目标分为规则类和非规则类两种,对非规则类前景目标如行人、动物等出现的断层分离现象设计前景目标区域纵向种子生长算法,对规则类前景目标如汽车轮船等设计区域内前景目标种子横纵双向连接以消除空洞、缺失的影响.结果 本文前景目标提取在富有挑战性干扰因素的复杂场景下体现出较高的鲁棒性,在数据库4组经典视频和山西太长高速公路2组视频中,动态背景有水流流动、树叶摇曳、摄像头轻微抖动、光照阴影,并从应用效果、前景目标定位的准确性以及前景目标检测的完整性3个角度对实验结果进行了分析,本文显著性前景目标提取算法取得了90.1％的平均准确率,88.7％的平均召回率以及89.4％的平均F值,均优于其他同类算法.结论 本文以快速定位显著性前景目标为前提,提出对稀疏低秩模型初筛选的图像进行并行种子识别和优化连接算法,实验数据的定性与定量分析结果表明,本文算法能够更快速地将前景目标与动态背景分离,并减小前景目标与背景之间的粘连情况,更有效地保留了原始图像中前景目标的结构信息.