Existing image segmentation algorithms face challenges related to low detection accuracy and a lack of specificity in crack detection. To address these challenges, this paper proposes an extended LG-Block module Extend-LG Block, which leverages a multi-scale feature fusion method. This module consists of multiple parallel dilated convolutions with different expansion rates. The number of branches and the expansion rate of dilated convolutions can be adjusted by parameters to change the size of its receptive field, and then extract and fuse crack features of different scales. By comparing the advantages and disadvantages of the network using a multi-scale feature fusion module in the deep layer and the network using a fixed scale structure for multi-scale feature fusion, a U-Net model with a variable scale structure named VS-UNet is proposed. The basic convolution Block in the UNet network is replaced by multiple Extend-LG blocks with different parameters. This structure performs multi-scale feature fusion in the shallow layer of the network, and the scale extracted by the multi-scale feature fusion module gradually decreases with the deepening of the network layer. This structure not only strengthens the detail feature extraction ability of the image while maintaining the original abstract feature extraction ability but also avoids the problem of increasing network parameters caused by the increase of convolution. Experiments are carried out on the DeepCrack dataset and CFD dataset. The results show that compared with the other two structures and methods, the proposed network with variable scale structure has higher detection accuracy and better segmentation effect for cracks of various sizes in visual experimental comparison. Finally, compared with other image segmentation algorithms, all indicators are improved to a certain extent compared with UNet, which proves the effectiveness of the improved network.
LG Block由两个分支组成,每个分支有一个3×3的空洞卷积,其扩张率分别为1和3,然后将两个扩张卷积操作的结果连接起来,以增强特征传播.然后,采用1×1卷积运算,在不改变特征图大小的情况下,加入非线性特征,实现各个不同尺度特征的融合.
LG Block仅增加一个膨胀率为3的空洞卷积,虽然扩大了感受野,但增加的尺度有限,在浅层网络中提取到的特征仍不够丰富,故需要在浅层增加提取的尺度.而受制于UNet网络中特征图尺寸大小随着池化层减少的特点,在网络深层无须增加大感受野卷积来提取大尺度信息.因此网络需要多种不同感受野的卷积块调节网络不同层的提取特征的尺度.针对上述问题,对LG Block进行扩展,扩展之后的模块Extend-LG Block的结构如图5所示.
采用不同多尺度特征提取方法的网络对比结果如表3所示,在P指标和R指标的对比上,3种多尺度网络结构在不同数据集上各有优劣,而在综合P指标与R指标的F1指标的对比上,本文提出的VS-UNet在两个数据集中均取得三者之中最高值.在DeepCrack数据集中,VS-UNet的F1指标比ASPP-UNet与FS-UNet分别高出3.68%,2.34%.在CFD数据集中,VS-UNet的F1指标比ASPP-UNet与FS-UNet分别高出1.20%,3.51%.在参数的对比上,VS-UNet比ASPP-UNet高1.02 M,比FS-UNet低3.24 M.
JIANGW B, LUOQ R, ZHANGX H .A review of concrete roads crack detection methods based on digital image[J].Journal of Xihua University (Natural Science Edition),2018,37(1):75-84.(in Chinese)
DENGL, CHUH H, LONGL Z,et al .Review of deep learning-based crack detection for civil infrastructures[J].China Journal of Highway and Transport,2023,36(2):1-21.(in Chinese)
[7]
KIRSCHKEK R, VELINSKYS A. Histogram-based approach for automated pavement-crack sensing[J]. Journal of Transportation Engineering,1992,118(5):700-710.
ZHANGA H, YUS S, ZHOUJ L .A local-threshold segment algorithm based on edge-detection[J]. Mini-micro Systems,2003,24(4):661-663.(in Chinese)
[10]
ZHANGL X, SHENJ K, ZHUB J. A research on an improved Unet-based concrete crack detection algorithm[J].Structural Health Monitoring,2021,20(4):1864-1879.
HUANGP, ZHENGQ, LIANGC .Overview of image segmentation methods[J].Journal of Wuhan University (Natural Science Edition),2020,66(6):519-531.(in Chinese)
[13]
SHELHAMERE, LONGJ, DARRELLT .Fully convolutional networks for semantic segmentation[C]//IEEE Transactions on Pattern Analysis and Machine Intelligence.IEEE,2017:640-651.
[14]
YANGX C, LIH, YUY T,et al .Automatic pixel‐level crack detection and measurement using fully convolutional network[J].Computer-Aided Civil and Infrastructure Engineering,2018, 33(12):1090-1109.
[15]
BADRINARAYANANV, KENDALLA, CIPOLLAR .SegNet:a deep convolutional encoder-decoder architecture for image segmentation[J].IEEE Transactions on Pattern Analysis and Machine Intelligence,2017,39(12):2481-2495.
[16]
RONNEBERGERO, FISCHERP, BROXT. U-Net:convolutional networks for biomedical image segmentation[M]//Lecture Notes in Computer Science.Cham:Springer International Publishing,2015:234-241.
GANL, XIEA R, YANY,et al .Crack segmentation of concrete surface based on improved U-Net[J].Journal of Chongqing University of Posts and Telecommunications (Natural Science Edition),2021,33(4):645-652.(in Chinese)
ZHUL X, WUR D, FUG P, et al. Segmenting banana images using the lightweight UNet of multi-scale serial dilated convolution[J]. Transactions of the Chinese Society of Agricultural Engineering,2022,38(13):194-201.(in Chinese)
SHIT T, GUOZ H, YANX,et al .Water body segmentation in remote sensing images based on multi-scale fusion attention module improved UNet[J].Chinese Journal of Liquid Crystals and Displays,2023,38(3):397-408.(in Chinese)
[25]
CHENL C, PAPANDREOUG, SCHROFFF, et al. Rethinking atrous convolution for semantic image segmentation[J]. arXiv prepriut arXiv:2017.
[26]
SONGW. PLU-Net: Extraction of multi-scale feature fusion[J]. 2023.
VAN HASSELTH, GUEZA, SILVERD. Deep reinforcement learning with double Q-learning[J].Proceedings of the AAAI Conference on Artificial Intelligence,2016,30(1):129-144.
[29]
WANGP Q, CHENP F, YUANY,et al .Understanding convolution for semantic segmentation[C]//2018 IEEE Winter Conference on Applications of Computer Vision (WACV). Lake Tahoe,NV,USA: IEEE,2018:1451-1460.
LIG Y, LIANGJ D, LIUY, et al. MFC-DeepLabV3+: a multi feature cascade fusion crack defect detection network model[J].Journal of Railway Science and Engineering, 2023, 20(4):1370-1381.(in Chinese)
XUX J, ZHANGX N .Crack detection of concrete bridges based digital image[J].Journal of Hunan University (Natural Sciences),2013,40(7):34-40.(in Chinese)
[34]
LIUY H, YAOJ, LUX H,et al .DeepCrack:a deep hierarchical feature learning architecture for crack segmentation[J].Neurocomputing,2019,338(03):139-153.
[35]
SHIY, CUIL M, QIZ Q,et al .Automatic road crack detection using random structured forests[J].IEEE Transactions on Intelligent Transportation Systems,2016,17(12):3434-3445.
[36]
LIUY, CHENGM M, HUX W,et al .Richer convolutional features for edge detection[C]//2017 IEEE Conference on Computer Vision and Pattern Recognition (CVPR).Honolulu,HI,USA: IEEE,2017:5872-5881.
[37]
JINGP, YUH Y, HUAZ H,et al .Road crack detection using deep neural network based on attention mechanism and residual structure[J].IEEE Access,2022,11:919-929.