Cargando…

Super-Resolution Reconstruction Method of Pavement Crack Images Based on an Improved Generative Adversarial Network

A super-resolution reconstruction approach based on an improved generative adversarial network is presented to overcome the huge disparities in image quality due to variable equipment and illumination conditions in the image-collecting stage of intelligent pavement detection. The nonlinear network o...

Descripción completa

Detalles Bibliográficos
Autores principales: Yuan, Bo, Sun, Zhaoyun, Pei, Lili, Li, Wei, Ding, Minghang, Hao, Xueli
Formato: Online Artículo Texto
Lenguaje:English
Publicado: MDPI 2022
Materias:
Acceso en línea:https://www.ncbi.nlm.nih.gov/pmc/articles/PMC9737262/
https://www.ncbi.nlm.nih.gov/pubmed/36501791
http://dx.doi.org/10.3390/s22239092
_version_ 1784847242031529984
author Yuan, Bo
Sun, Zhaoyun
Pei, Lili
Li, Wei
Ding, Minghang
Hao, Xueli
author_facet Yuan, Bo
Sun, Zhaoyun
Pei, Lili
Li, Wei
Ding, Minghang
Hao, Xueli
author_sort Yuan, Bo
collection PubMed
description A super-resolution reconstruction approach based on an improved generative adversarial network is presented to overcome the huge disparities in image quality due to variable equipment and illumination conditions in the image-collecting stage of intelligent pavement detection. The nonlinear network of the generator is first improved, and the Residual Dense Block (RDB) is created to serve as Batch Normalization (BN). The Attention Module is then formed by combining the RDB, Gated Recurrent Unit (GRU), and Conv Layer. Finally, a loss function based on the L1 norm is utilized to replace the original loss function. The experimental findings demonstrate that the self-built pavement crack dataset’s Peak Signal-to-Noise Ratio (PSNR) and Structural Similarity (SSIM) of the reconstructed images reach 29.21 dB and 0.854, respectively. The results improved compared to the Set5, Set14, and BSD100 datasets. Additionally, by employing Faster-RCNN and a Fully Convolutional Network (FCN), the effects of image reconstruction on detection and segmentation are confirmed. The findings indicate that the segmentation results’ F1 is enhanced by 0.012 to 0.737 and the detection results’ confidence is increased by 0.031 to 0.9102 when compared to state-of-the-art methods. It has a significant engineering application value and can successfully increase pavement crack-detecting accuracy.
format Online
Article
Text
id pubmed-9737262
institution National Center for Biotechnology Information
language English
publishDate 2022
publisher MDPI
record_format MEDLINE/PubMed
spelling pubmed-97372622022-12-11 Super-Resolution Reconstruction Method of Pavement Crack Images Based on an Improved Generative Adversarial Network Yuan, Bo Sun, Zhaoyun Pei, Lili Li, Wei Ding, Minghang Hao, Xueli Sensors (Basel) Article A super-resolution reconstruction approach based on an improved generative adversarial network is presented to overcome the huge disparities in image quality due to variable equipment and illumination conditions in the image-collecting stage of intelligent pavement detection. The nonlinear network of the generator is first improved, and the Residual Dense Block (RDB) is created to serve as Batch Normalization (BN). The Attention Module is then formed by combining the RDB, Gated Recurrent Unit (GRU), and Conv Layer. Finally, a loss function based on the L1 norm is utilized to replace the original loss function. The experimental findings demonstrate that the self-built pavement crack dataset’s Peak Signal-to-Noise Ratio (PSNR) and Structural Similarity (SSIM) of the reconstructed images reach 29.21 dB and 0.854, respectively. The results improved compared to the Set5, Set14, and BSD100 datasets. Additionally, by employing Faster-RCNN and a Fully Convolutional Network (FCN), the effects of image reconstruction on detection and segmentation are confirmed. The findings indicate that the segmentation results’ F1 is enhanced by 0.012 to 0.737 and the detection results’ confidence is increased by 0.031 to 0.9102 when compared to state-of-the-art methods. It has a significant engineering application value and can successfully increase pavement crack-detecting accuracy. MDPI 2022-11-23 /pmc/articles/PMC9737262/ /pubmed/36501791 http://dx.doi.org/10.3390/s22239092 Text en © 2022 by the authors. https://creativecommons.org/licenses/by/4.0/Licensee MDPI, Basel, Switzerland. This article is an open access article distributed under the terms and conditions of the Creative Commons Attribution (CC BY) license (https://creativecommons.org/licenses/by/4.0/).
spellingShingle Article
Yuan, Bo
Sun, Zhaoyun
Pei, Lili
Li, Wei
Ding, Minghang
Hao, Xueli
Super-Resolution Reconstruction Method of Pavement Crack Images Based on an Improved Generative Adversarial Network
title Super-Resolution Reconstruction Method of Pavement Crack Images Based on an Improved Generative Adversarial Network
title_full Super-Resolution Reconstruction Method of Pavement Crack Images Based on an Improved Generative Adversarial Network
title_fullStr Super-Resolution Reconstruction Method of Pavement Crack Images Based on an Improved Generative Adversarial Network
title_full_unstemmed Super-Resolution Reconstruction Method of Pavement Crack Images Based on an Improved Generative Adversarial Network
title_short Super-Resolution Reconstruction Method of Pavement Crack Images Based on an Improved Generative Adversarial Network
title_sort super-resolution reconstruction method of pavement crack images based on an improved generative adversarial network
topic Article
url https://www.ncbi.nlm.nih.gov/pmc/articles/PMC9737262/
https://www.ncbi.nlm.nih.gov/pubmed/36501791
http://dx.doi.org/10.3390/s22239092
work_keys_str_mv AT yuanbo superresolutionreconstructionmethodofpavementcrackimagesbasedonanimprovedgenerativeadversarialnetwork
AT sunzhaoyun superresolutionreconstructionmethodofpavementcrackimagesbasedonanimprovedgenerativeadversarialnetwork
AT peilili superresolutionreconstructionmethodofpavementcrackimagesbasedonanimprovedgenerativeadversarialnetwork
AT liwei superresolutionreconstructionmethodofpavementcrackimagesbasedonanimprovedgenerativeadversarialnetwork
AT dingminghang superresolutionreconstructionmethodofpavementcrackimagesbasedonanimprovedgenerativeadversarialnetwork
AT haoxueli superresolutionreconstructionmethodofpavementcrackimagesbasedonanimprovedgenerativeadversarialnetwork