A platform for research: civil engineering, architecture and urbanism
A platform for research: civil engineering, architecture and urbanism
Deep CNN-based concrete cracks identification and quantification using image processing techniques
In evaluating the quality of concrete structures, cracks play a significant role in determining the structure’s safety, application, and durability. Manual quality checks are more susceptible to human error and take longer to complete. As a result, utilizing computer algorithms to see fissures and identify problems in concrete structures is now the recommended method. The current research focuses on developing the deep convolutional neural network (CNN) model to identify the crack/defect. Subsequently, the quantification of cracks has been determined using the image processing method by building the Python source code. The suggested CNN architecture is developed using the VGG16 model and was trained on a pre-build-up dataset of 40,000 images of the pixel size of 227 × 227; the results are obtained with an accuracy of 94.6%. Manually 280 images are collected irrespective of pixel size and distance between the surface and the source with known crack width are utilized in an image processing technique. The comparison between the test results of this technique and the measurements obtained by actual physical use of a crack microscope is considered to ensure the effectiveness of the planned digital image processing (DIP) systems. The quantification results obtained through Image processing with the percentage of accuracy vary from 65 to 98% by comparison with the measurements gained by the real physical way.
Deep CNN-based concrete cracks identification and quantification using image processing techniques
In evaluating the quality of concrete structures, cracks play a significant role in determining the structure’s safety, application, and durability. Manual quality checks are more susceptible to human error and take longer to complete. As a result, utilizing computer algorithms to see fissures and identify problems in concrete structures is now the recommended method. The current research focuses on developing the deep convolutional neural network (CNN) model to identify the crack/defect. Subsequently, the quantification of cracks has been determined using the image processing method by building the Python source code. The suggested CNN architecture is developed using the VGG16 model and was trained on a pre-build-up dataset of 40,000 images of the pixel size of 227 × 227; the results are obtained with an accuracy of 94.6%. Manually 280 images are collected irrespective of pixel size and distance between the surface and the source with known crack width are utilized in an image processing technique. The comparison between the test results of this technique and the measurements obtained by actual physical use of a crack microscope is considered to ensure the effectiveness of the planned digital image processing (DIP) systems. The quantification results obtained through Image processing with the percentage of accuracy vary from 65 to 98% by comparison with the measurements gained by the real physical way.
Deep CNN-based concrete cracks identification and quantification using image processing techniques
Asian J Civ Eng
Gonthina, Madhuri (author) / Chamata, Renuka (author) / Duppalapudi, Jhanshi (author) / Lute, Venkat (author)
Asian Journal of Civil Engineering ; 24 ; 727-740
2023-04-01
14 pages
Article (Journal)
Electronic Resource
English
Detection and measurement of concrete cracks by image processing techniques
| British Library Conference Proceedings | 2006
Identification and quantification of cracks in concrete by optical fluorescent microscopy
| British Library Online Contents | 2006
Identification and quantification of cracks in concrete by optical fluorescent microscopy
| Online Contents | 2006
Identification and quantification of cracks in concrete by optical fluorescent microscopy
| British Library Conference Proceedings | 2006