Issue 65

L. Wang, Frattura ed Integrità Strutturale, 65 (2023) 289-299; DOI: 10.3221/IGF-ESIS.65.19

C ONCLUSION racks in concrete infrastructure are crucial indicators of structural damage. They severely affect structural integrity and functionality. In this research, we proposed an automatic decision-making system CrackSN based on deep learning to inspect cracks in captured digital images. The experimental results of our model provide the following conclusions: 1. Through the training and validation process, the CrackSN system automatically learns the features from manually annotated image dataset and successfully predicts 97.3% of the cracked patches in the test datasets. The presented Adam-SqueezeNet deep learning based CrackSN system exhibits superior performance in crack detection compared with state-of-the-art models. 2. The deep Adam-SqueezeNet model with a small structural size and superior performance is preferred in embedded applications for real-time crack or damage detection in concrete infrastructure for fast damage detection and decision making of maintenance. 3. The performance of our model can be further improved to pixel-level accuracy by incorporating up-sampling layers or other deep learning algorithm. Training the presented model on a wider variety of image datasets with realistic scenarios is also necessary for practical engineering applications. A CKNOWLEDGMENT he financial support from the University Nature Sciences Research Program of Anhui Province (KJ2021A1001) and the Academic Visit Program of Anhui Province (gxgnfx2022054) is greatly appreciated. C

T

N OMENCLATURE Symbol

Meaning

Symbol

Meaning feature map max pool padding length image patch output of softmax iteration number loss function learning rate

x

matrix of input image weight of convolution filter

y z

w

s

f

step size of stride image dataset

S L m

P q

class label of image patch parameter gradient

v

squared parameter gradient

β 1 , β 2

l

decay rates

θ γ

E (θ)

parameter vector

α

rotational angle of image patch

R EFERENCE [1] Sony, S., Laventure, S., Sadhu, A., (2019). A literature review of next-generation smart sensing technology in structural health monitoring. Struct. Control Health Monit., 26(3), e2321. DOI: 10.1002/stc.2321 [2] Ye, X. W., Jin, T., Yun, C. B., (2019). A review on deep learning-based structural health monitoring of civil infrastructures. Smart Struct. Syst., 24(5), pp. 567-585. DOI: 10.12989/sss.2019.24.5.567. [3] Fei, Y., Wang, K. C., Zhang, A., Chen, C., Li, J. Q., Liu, Y., Li, B., (2019). Pixel-level cracking detection on 3D asphalt pavement images through deep-learning-based CrackNet-V. IEEE trans. Intell. Transp. Syst., 21(1), pp. 273-284. DOI: 10.1109/TITS.2019.2891167. [4] Alipour, M., Harris, D. K., and Miller, G. R., (2019). Robust pixel-level crack detection using deep fully convolutional neural networks. J. Comput. Civ. Eng., 33(6), 04019040. DOI: 10.1061/(ASCE)CP.1943-5487.0000854. [5] Ni, F., Zhang, J., Chen, Z., (2019). Pixel ‐ level crack delineation in images with convolutional feature fusion. Struct. Control Health Monit., 26(1), e2286. DOI: 10.1002/stc.2286.

298