Please wait a minute...
浙江大学学报(工学版)
Journal of ZheJiang University (Engineering Science)  2023, Vol. 57 Issue (10): 2094-2105    DOI: 10.3785/j.issn.1008-973X.2023.10.018
    
Intelligent identification of asphalt pavement cracks based on semantic segmentation
Yan-ze YANG1(),Meng WANG1,*(),Cheng LIU2,Hui-tong XU1,Xiao-yue ZHANG1
1. School of Civil Engineering, Beijing Jiaotong University, Beijing 100044, China
2. China Road Transportation Verification and Inspection Hi-Tech Co Ltd., Beijing 100088, China
Download: HTML     PDF(1864KB) HTML
Export: BibTeX | EndNote (RIS)      

Abstract  

An intelligent method of asphalt pavement crack recognition based on semantic segmentation was proposed, solving the shortcomings of traditional manual inspection of asphalt pavement, such as low efficiency and lack of objectivity. Considering the effects of data set size, algorithm type, network type and depth, and loss function type, the optimal crack intelligent identification scheme and corresponding model were proposed for both large and small scale data sets through the comparative study of 22 semantic segmentation models. Based on the asphalt pavement of sixth ring road in Beijing, the crack segmentation dataset R-Crack was established. The proposed intelligent identification scheme was verified and the crack parameters were automatically quantified. Results showed that the highest detection accuracy reached 83.45%. The average errors of crack length and width were 2.84% and 2.39% respectively by comparing the calculation results of crack parameters obtained through manual and automatic detection methods, The proposed intelligent recognition scheme provided a basis for the intelligent detection practice of asphalt pavement cracks in the expressway and other scenes.

Key words： asphalt pavement inspection      intersection over union      semantic segmentation      crack identification      convolution neural network     
Received: 05 December 2022      Published: 18 October 2023
CLC:  U 416.217  
Fund:  中央高校基本科研业务费专项资金资助项目（2022YJS071）；北京市科技新星计划资助项目（20220484103）；北京市自然科学基金资助项目（8222027）
Corresponding Authors: Meng WANG     E-mail: 21121139@bjtu.edu.cn;wangmeng@bjtu.edu.cn
Service
E-mail this article
Add to my bookshelf
Add to citation manager
E-mail Alert
RSS
Articles by authors
Yan-ze YANG
Meng WANG
Cheng LIU
Hui-tong XU
Xiao-yue ZHANG
Cite this article:

Yan-ze YANG,Meng WANG,Cheng LIU,Hui-tong XU,Xiao-yue ZHANG. Intelligent identification of asphalt pavement cracks based on semantic segmentation. Journal of ZheJiang University (Engineering Science), 2023, 57(10): 2094-2105.

URL:

https://www.zjujournals.com/eng/10.3785/j.issn.1008-973X.2023.10.018     OR     https://www.zjujournals.com/eng/Y2023/V57/I10/2094


基于语义分割的沥青路面裂缝智能识别

针对传统的沥青路面人工检测效率低、缺乏客观性的弊端，提出基于语义分割的沥青路面裂缝智能识别方法. 综合考虑数据集规模、算法种类、网络种类及深度、损失函数类型的影响，对 22 个语义分割模型开展对比研究，提出适用于较大、较小规模数据集的优选裂缝智能识别方案及对应模型. 基于北京六环高速公路沥青路面，建立裂缝分割数据集R-Crack，对提出的智能识别方案进行应用检验，并自动量化裂缝参数. 结果表明：检测准确率最高达到83.45%，通过对比人工及自动化检测方式获得的裂缝参数计算结果，裂缝长度和宽度平均误差分别为2.84%和2.39%，提出的智能识别方案为高速公路等场景下沥青路面裂缝的智能检测实践提供依据.


关键词： 沥青路面检测,  交并比,  语义分割,  裂缝识别,  卷积神经网络 
时间 网络 特点及优势 存在问题
2014 FCN 传统CNN的全连接层换成卷积层，可以对图像每个像素预测，实现任意分辨率图像的输入. 结果准确率不高，忽略像素之间的关系，缺乏空间一致性
2015 U-Net 采用编码器解码器结构、跨层连接编码解码过程、网络结构小、模型简洁. 速度慢，检测区域重叠，定位准确性和上下文信息不兼得
2016 PSPNet 在FCN基础上，提出金字塔池化模块，融合局部和全局获取上下文信息，提取特征多样化，包含不同尺度信息. 分割目标的边界信息部分易丢失
2017 DeepLabV3 改进金字塔池化模块，使用全局平均池化结构，强调全局特征. 分割速度慢，对小尺寸物体分割效果不明显
2018 DeepLabV3+ 新的编码-解码结构实现多尺度上下文信息探索，优化运行速度. —
Tab.1 Summary of common semantic segmentation algorithms
数据集名称 N R Bit
CRACK500[26] 2 244 640 × 360 24
GAPS384[27] 509 540 × 640
540 × 440 		24
Tab.2 Basic information of training crack datasets
Fig.1 Example of training crack datasets
数据集 训练集数 验证集数 测试集数
CRACK500 1 753 430 123
GAPS384 407 41 123
Tab.3 Dataset partitioning of semantic segmentation comparison experiments
Fig.2 Comparative schemes for semantic segmentation models
Fig.3 Flowchart of semantic segmentation model training
指标全称 指标说明 计算公式
IOU 交并比：真实分割与系统预测分割结果两个几何之间的比值，IOU越大，表明模型预测到的裂缝与实际图片中的裂缝重合度越高. ${\rm{IOU}} = \dfrac{ {{\rm{TP}}} }{ {({\rm{TP}}+{\rm{FP}}+{\rm{FN}})} }.$
Accuracy 准确率：预测正确的像素数量占实际总像素数量的比例，Accuracy值越大，表明模型在像素预测的正确性上效果更好. ${\rm{Acc}} = \dfrac{ {{\rm{TP}}+{\rm{TN}}} }{ {{\rm{TP}}+{\rm{TN}}+{\rm{FP}}+{\rm{FN}}} }.$
Precision 精确率：针对预测结果而言的，用于衡量模型检测的有多准，一类实际像素个数占模型预测为该类像素的比例，值越大，表明模型在该类的所有预测结果中正确预测的可能性越高. ${\rm{Pr}} = \dfrac{ {{\rm{TP}}} }{ {{\rm{TP}}+{\rm{FP}}} }.$
Recall 召回率：针对原样本而言的，用于衡量模型预测结果有多全面的指标；其含义是在实际为正的样本中被预测为正样本的概率. ${\rm{Re}}=\dfrac{{\rm{TP}}}{{\rm{TP}}+{\rm{FN}}}.$
Fscore F1分数：它以两者同样重要的权重将召回率和精确率加权调和平均，该值越大，表明被正确识别的裂缝像素越多，分割结果越精确，效果越好. ${\rm{F}}1 = \dfrac{ {2{\rm{Pr}}\times {\rm{Re}}} }{ {{\rm{Pr}}+{\rm{Re}}} }.$
Loss 损失函数值：损失函数是表现深度学习模型预测和实际数据差距程度的函数，其值越小，说明模型的鲁棒性越好. ——
Tab.4 Summary of evaluation indicators for experiment
模型序号 算法 网络 损失函数 占用空间/MB 训练时长/h
模型1（M1） U-Net R101 CROSS 423 3.12
模型2（M2） DeepLabV3 R101 CROSS 665 4.56
模型3（M3） PSPNet R101 CROSS 519 3.43
模型4（M4） DeepLabV3+ R101 CROSS 478 3.53
模型5（M5） DeepLabV3 R101 FOCAL 665 3.79
模型6（M6） PSPNet R101 FOCAL 519 2.72
模型7（M7） DeepLabV3 R101 DICE 665 3.78
模型8（M8） PSPNet R101 DICE 519 2.71
模型9（M9） U-Net R101 CROSS 423 11.68
模型10（M10） DeepLabV3 R101 CROSS 665 20.40
模型11（M11） PSPNet R101 CROSS 519 15.80
模型12（M12） DeepLabV3+ R101 CROSS 478 14.21
模型13（M13） DeepLabV3 R101 FOCAL 665 18.29
模型14（M14） PSPNet R101 FOCAL 519 12.98
模型15（M15） DeepLabV3 R101 DICE 665 18.23
模型16（M16） PSPNet R101 DICE 519 15.86
模型17（M17） DeepLabV3 R50 CROSS 518 13.40
模型18（M18） DeepLabV3 R18 CROSS 106 4.80
模型19（M19） DeepLabV3 MV2 CROSS 142 4.10
模型20（M20） PSPNet R50 CROSS 373 8.20
模型21（M21） PSPNet R18 CROSS 97.6 4.90
模型22（M22） PSPNet MV2 CROSS 104 3.90
Tab.5 Basic parameters of training model
Fig.4 Accuracy and loss curves of different algorithm models
分类 模型序号 IOU/% Acc/% F1/% FPS（帧/s）
小数据集 M1_U-Net 57.78 59.16 0.62 0.45
M2_Deeplab V3 63.91 65.62 0.71 0.87
M3_PSPNet 65.51 67.18 0.73 0.85
M4_ Deeplab V3+ 65.27 67.29 0.73 0.69
大数据集 M9_ U-Net 69.59 85.09 0.79 0.87
M10_ Deeplab V3 75.60 86.30 0.84 0.56
M11_ PSPNet 75.12 86.76 0.84 0.71
M12_ Deeplab V3+ 75.70 86.66 0.85 0.63
Tab.6 Test results of different algorithm models
Fig.5 Area comparison of false identification (FN) and missing identification (FP) of model segmentation results
Fig.6 Model 9~12 visualization of crack segmentation results
Fig.7 Accuracy curves of different loss function models
分类 模型序号 IOU/% Acc/% F1/% FPS（帧/s）
小数据集 M2_CROSS 63.91 65.62 0.71 0.87
M5_FOCAL 61.70 63.33 0.68 0.87
M7_DICE 58.13 59.63 0.64 0.87
M3_CROSS 65.51 67.18 0.73 0.85
M6_FOCAL 63.29 65.61 0.70 0.82
M8_DICE 58.34 59.75 0.63 0.85
大数据集 M10_CROSS 75.90 86.30 0.85 0.56
M13_FOCAL 77.41 85.38 0.86 0.55
M15_DICE 71.60 84.76 0.81 0.55
M11_CROSS 75.12 86.76 0.84 0.71
M14_FOCAL 74.13 85.31 0.83 0.67
M16_DICE 69.94 84.71 0.79 0.67
Tab.7 Test results of different loss function models
Fig.8 Accuracy and loss curves of different network structures
模型序号 IOU/% Acc/% F1/% FPS/（帧·s?1）
M10_R101 75.90 86.30 0.85 0.56
M17_R50 77.96 86.72 0.87 0.62
M18_R18 72.72 85.53 0.82 2.67
M19_MV2 75.67 85.59 0.85 3.05
M11_R101 75.12 86.76 0.84 0.71
M20_R50 76.20 85.69 0.85 1.03
M21_R18 74.75 86.23 0.84 2.67
M22_MV2 71.71 82.09 0.81 4.56
Tab.8 Test results of different network structure models
Fig.9 Optimal scheme of intelligent detection model for asphalt pavement cracks
Fig.10 Flowchart of integrated solution for automatic identification and quantification of asphalt pavement cracks
Fig.11 Flowchart of asphalt pavement crack segmentation dataset R-Crack construction
Fig.12 Example of "R-Crack" dataset annotation
Fig.13 Sketch map of crack skeleton extraction
Fig.14 Schematic diagram of calculation principle of maximum inscribed circle crack width
数据集名称 模型名称 IOU/% Acc/% F1/% FPS（帧/s）
R-Crack DeepLabV3+
_R101_CROSS 		79.56 83.45 0.85 0.65
R-Crack PSPNet_R101_CROSS 78.49 82.14 0.81 0.39
Tab.9 Application test results of optimized model
Fig.15 Crack detection results of sixth ring road in Beijing
Fig.16 Extraction results of crack skeleton
[1]   交通运输部 2020年全国收费公路统计公报[J]. 交通财会, 2021, (11): 93- 96
Ministry of Transport 2020 national toll road statistics bulletin[J]. Transportation Finance and Accounting, 2021, (11): 93- 96
[2]   徐鹏, 祝轩, 姚丁, 等 沥青路面养护智能检测与决策综述[J]. 中南大学学报: 自然科学版, 2021, 52 (7): 2099- 2117
XU Peng, ZHU Xuan, YAO Ding, et al Review on intelligent detection and decision-making of asphalt pavement maintenance[J]. Journal of Central South University: Science and Technology, 2021, 52 (7): 2099- 2117
[3]   马建, 赵祥模, 贺拴海, 等 路面检测技术综述[J]. 交通运输工程学报, 2017, 17 (5): 121- 137
MA Jian, ZHAO Xiang-mo, HE Shuan-hai, et al Review of pavement detection technology[J]. Journal of Traffic and Transportation Engineering, 2017, 17 (5): 121- 137
doi: 10.3969/j.issn.1671-1637.2017.05.012
[4]   交通运输部公路研究院. 公路技术状况评定标准: JTG 5210-2018 [S]. 北京: 人民交通出版社, 2018: 17-21.
[5]   交通运输部公路研究院. 公路路面技术状况自动化检测规程: JTGT E61-2014 [S]. 北京: 人民交通出版社, 2014: 11-14.
[6]   黎蔚, 朱平哲 沥青路面裂缝图像检测算法研究[J]. 计算机工程与应用, 2012, 48 (19): 163- 166
LI Wei, ZHU Ping-zhe Image detection algorithm research for asphalt pavement crack[J]. Computer Engineering and Applications, 2012, 48 (19): 163- 166
doi: 10.3778/j.issn.1002-8331.2012.19.037
[7]   翁飘, 陆彦辉, 齐宪标, 等. 基于改进的全卷积神经网络的路面裂缝分割技术[J]. 计算机工程与应用, 2019, 55(16): 235-239.
WENG Piao, LU Yan-hui, QI Xian-biao, et al. Pavement crack segmentation technology based on improved fully convolutional networks [J]. Computer Engineering and Applications. 2019, 55(16): 235-239.
[8]   赵颖. 基于卷积神经网络的沥青路面裂缝图像筛选与分割[D]. 石家庄: 石家庄铁道大学, 2021: 9-13.
ZHAO Ying. Pavement crack segmentation technology based on improved fully convolutional networks [D]. Shijiazhuang: Shijiazhuang Railway University, 2021: 9-13.
[9]   唐由之. 基于卷积神经网络的沥青路面裂缝智能识别算法研究[D]. 成都: 西南交通大学, 2021: 13-22.
TANG You-zhi. Research on intelligent cracks detection algorithm of asphalt pavement based on convolutional neural network [D]. Chengdu: Southwest Jiaotong University, 2021: 13-22.
[10]   HSIEH Y A, TSAI Y J Machine learning for crack detection: review and model performance comparison[J]. Journal of Computing in Civil Engineering, 2021, 34 (5): 04020038
[11]   田萱, 王亮, 丁琪 基于深度学习的图像语义分割方法综述[J]. 软件学报, 2019, 30 (2): 440- 468
TIAN Xuan, WANG Liang, DING Qi Overview of image semantic segmentation methods based on deep learning[J]. Journal of Software, 2019, 30 (2): 440- 468
doi: 10.13328/j.cnki.jos.005659
[12]   李刚, 高振阳, 张新春, 等 改进的全局卷积网络在路面裂缝检测中的应用[J]. 激光与光电子学进展, 2020, 57 (8): 111- 119
LI Gang, GAO Zhen-yang, ZHANG Xin-chun, et al Application of improved global convolution network in pavement crack detection[J]. Progress in Laser and Optoelectronics, 2020, 57 (8): 111- 119
[13]   陈泽斌, 罗文婷, 李林. 基于改进U-net模型的路面裂缝智能识别[J]. 数据采集与处理. 2020, 35(2): 260-269.
CHEN Ze-bin, LUO Wen-ting, LI Lin. Automatic identification of pavement crack using improved U-net model [J]. Journal of Data Acquisition and Processing, 2020, 35(2): 260-269.
[14]   阙云, 季雪, 蒋子平, 等. GAN数据增强下路面裂缝语义分割算法[EB/OL]. (2022-07-08) [2022-12-04]. http://doi.org/10.13229/j.cnki.jdxbgxb20220003.
QUE Yun, JI Xue, JIANG Zi-ping, et al. Semantic segmentation algorithm of pavement cracks based on GAN data augmentation [EB/OL]. (2022-07-08)[2022-12-04]. http://doi.org/10.13229/j.cnki.jdxbgxb20220003.
[15]   ZHANG A, WANGK C P, FEI Y, et al Automated pixe mmevel pavement crack detection on 3D asphalt surfaces with a recurrent neural network[J]. Computer Aided Civil and Infrastruc Ture Engineering, 2018, 34 (3): 213- 229
[16]   XIANG X, ZHANG Y, SADDIK ABDULMOTALEB S Pavement crack detection network based on pyramid structure and attention mechanism[J]. IET Image Processing, 2020, 14 (8): 1580- 1586
doi: 10.1049/iet-ipr.2019.0973
[17]   YANG F, LEI Z, SIJIA Y, et al Feature pyramid and hierarchical boosting network for pavement crack detection[J]. IEEE Transactions on Intelligent Transportation Systems, 2020, (4): 1525- 1535
[18]   韩晓健, 赵志成 基于计算机视觉技术的结构表面裂缝检测方法研究[J]. 建筑结构学报, 2018, 39 (增1): 418- 427
HAN Xiao-jian, ZHAO Zhi-cheng Structural surface crack detection method based on computer vision technology[J]. Journal of Building Structures, 2018, 39 (增1): 418- 427
doi: 10.14006/j.jzjgxb.2018.S1.055
[19]   周颖, 刘彤 基于计算机视觉的混凝土裂缝识别[J]. 同济大学学报: 自然科学版, 2019, 47 (9): 1277- 1285
ZHOU Ying, LIU Tong Computer vision-based crack detection and measurement on concrete structure[J]. Journal of Tongji University: Natural Science, 2019, 47 (9): 1277- 1285
[20]   孟诗乔, 张啸天, 乔甦阳, 等 基于深度学习的网格优化裂缝检测模型研究[J]. 建筑结构学报, 2020, 41 (增2): 404- 410
MENG Shi-qiao, ZHAO Xiao-tian, QIAO Su-yang, et al Research on grid optimized crack detection model based on deep learning[J]. Journal of Building Structures, 2020, 41 (增2): 404- 410
doi: 10.14006/j.jzjgxb.2020.S2.0045
[21]   丁威, 俞珂, 舒江鹏 基于深度学习和无人机的混凝土结构裂缝检测方法[J]. 土木工程学报, 2021, 54 (增1): 1- 12
DING Wei, YU Ke, SHU Jiang-peng Method for detecting cracks in concrete structures based on deep learning and UAV[J]. Journal of Civil Engineering, 2021, 54 (增1): 1- 12
[22]   CSURKA G, PERRONNIN F An efficient approach to semantic segmentation[J]. International Journal of Computer Vision, 2011, 95 (2): 198- 212
doi: 10.1007/s11263-010-0344-8
[23]   EVERINGHAM M, ESLAMI S M, GOOL L V, et al The pascal visual object classes challenge: a retrospective[J]. International Journal of Computer Vision, 2015, 111 (1): 98- 136
doi: 10.1007/s11263-014-0733-5
[24]   LIN T Y, MAIRE M, BELONGIE S, et al. Microsoft coco: common objects in context [C]// Computer Vision–ECCV. Zürich: SIP, 2014: 740-755.
[25]   CORDTS M, OMRAN M, RAMOS S, et al. The cityscapes dataset for semantic urban scene understanding [C]// IEEE Conference on Computer Vision and Pattern Recognition (CVPR). Las Vegas: IEEE, 2016: 2039-2047.
[26]   EISENBACH M, STRICKER R, SEICHTER D, et al. How to get pavement distress detection ready for deep learning? a systematic approach [C]// International Joint Conference on Neural Networks. Anchorage: IEEE, 2017: 2039−2047.
[27]   LEON B, FRANK E C, JORGE N Optimization methods for large-scale machine learning[J]. Siam Review, 2018, 60 (2): 223- 311
[28]   任凤雷, 何昕, 魏仲慧, 等 基于DeepLabV3+与超像素优化的语义分割[J]. 光学精密工程, 2019, 27 (12): 2722- 2729
REN Feng-lei, HE Xin, WEI Zhong-hui Semantic segmentation based on DeepLabV3+ and super pixel optimization[J]. Optics and Precision Engineering, 2019, 27 (12): 2722- 2729
[29]   于桐, 吴文瑾, 刘海江, 等 基于改进U-Net网络与联合损失函数的海南自然保护区高分辨率遥感变化检测模型[J]. 中国环境监测, 2021, 37 (5): 194- 200
YU Tong, WU Wen-jin, LIU Jiang-hai, et al Remote sensing change detection model of Hainan nature reserves based on improved U-Net and joint loss function[J]. Environmental Monitoring in China, 2021, 37 (5): 194- 200
[30]   LIN T Y, GOYAL P, GIRSHICK R, et al Focal loss for dense object detection[J]. IEEE Transactions on Pattern Analysis and Machine Intelligence, 2020, 42 (2): 318- 327
[31]   HE K, ZHANG X, REN S, et al. Deep residual learning for image recognition [C]// IEEE Conference on Computer Vision and Pattern Recognition. Las Vegas: IEEE, 2016: 770−778.
[32]   SANDLER M, HOCARD A, ZHU M, et al. MobileNetV2: inverted residuals and linear bottlenecks [C]// IEEE/CVF Conference on Computer Vision and Pattern Recognition. Salt Lake City: IEEE, 2018: 4510−4520.
[33]   张晋赫, 秦育罗, 张在岩, 等 复杂场景下农村道路裂缝分割方法[J]. 测绘通报, 2022, 66 (5): 74- 78
ZHANG Jin-he, QIN Yu-luo, ZHANG Zai-yan, et al Rural road crack segmentation method in complex scene[J]. Bulletin of Surveying and Mapping, 2022, 66 (5): 74- 78
[34]   汪德佳. 基于计算机视觉方法的古建筑变形监测[D]. 北京: 北京交通大学, 2022: 67−68.
Wang De-jia. Deformation monitoring of ancient buildings based on computer vision methods [D]. Beijing: Beijing Jiaotong University, 2022: 67−68.
[35]   刘宇飞. 基于模型修正与图像处理的多尺度结构损伤识别[D]. 北京: 清华大学, 2015: 188−189.
LIU Yu-fei. Multi-scale structural damage assessment based on model updating and image processing [D]. Beijing: Tsinghua University, 2015: 188−189.
[1] Hao-ran GUO,Ji-chang GUO,Yu-dong WANG. Lightweight semantic segmentation network for underwater image[J]. Journal of ZheJiang University (Engineering Science), 2023, 57(7): 1278-1286.
[2] Chun-juan LIU,Ze QIAO,Hao-wen YAN,Xiao-suo WU,Jia-wei WANG,Yu-qiang XIN. Semantic segmentation network for remote sensing image based on multi-scale mutual attention[J]. Journal of ZheJiang University (Engineering Science), 2023, 57(7): 1335-1344.
[3] Hai-bo ZHANG,Lei CAI,Jun-ping REN,Ru-yan WANG,Fu LIU. Efficient and adaptive semantic segmentation network based on Transformer[J]. Journal of ZheJiang University (Engineering Science), 2023, 57(6): 1205-1214.
[4] Yu-xiang WANG,Zhi-wei ZHONG,Peng-cheng XIA,Yi-xiang HUANG,Cheng-liang LIU. Compound fault decoupling diagnosis method based on improved Transformer[J]. Journal of ZheJiang University (Engineering Science), 2023, 57(5): 855-864.
[5] Qing ZHAO,Xue-ying ZHANG,Gui-jun CHEN,Jing ZHANG. EEG and fNIRS emotion recognition based on modality attention graph convolution feature fusion[J]. Journal of ZheJiang University (Engineering Science), 2023, 57(10): 1987-1997.
[6] Wei SUN,Heng LIU,Jian-feng TAO,Hao SUN,Cheng-liang LIU. IndRNN-1DLCNN based fault diagnosis of independent metering valve-controlled hydraulic cylinder system[J]. Journal of ZheJiang University (Engineering Science), 2023, 57(10): 2028-2041.
[7] Ze-kang WU,Shan ZHAO,Hong-wei LI,Yi-rui JIANG. Spatial global context information network for semantic segmentation of remote sensing image[J]. Journal of ZheJiang University (Engineering Science), 2022, 56(4): 795-802.
[8] Deng-wen ZHOU,Jin-yue TIAN,Lu-yao MA,Xiu-xiu SUN. Lightweight image semantic segmentation based on multi-level feature cascaded network[J]. Journal of ZheJiang University (Engineering Science), 2020, 54(8): 1516-1524.
[9] Xu YAN,Xiao-liang FAN,Chuan-pan ZHENG,Yu ZANG,Cheng WANG,Ming CHENG,Long-biao CHEN. Urban traffic flow prediction algorithm based on graph convolutional neural networks[J]. Journal of ZheJiang University (Engineering Science), 2020, 54(6): 1147-1155.
[10] ZHANG Jun-yang, GUO Yang, HU Xiao. Parallel computing method for two-dimensional matrix convolution[J]. Journal of ZheJiang University (Engineering Science), 2018, 52(3): 515-523.
[11] ZHANG Lin, CHENG Hua, FANG Yi-quan. CNN-based link representation and prediction method[J]. Journal of ZheJiang University (Engineering Science), 2018, 52(3): 552-559.
[12] ZHANG Peng-wei, LIU Zhen-tao, LI Jian-feng, LI Jing-lu, YU Xiao-li. Fatigue crack identification method based on relative
change of strain in engine block[J]. Journal of ZheJiang University (Engineering Science), 2012, 46(5): 948-953.