Research on Automatic Crack Detection for Concrete Infrastructures Using Image Processing and Deep Learning

Automatic crack detection is a critical task in the generation of a crack map for existing concrete infrastructure inspection. This paper describes an automatic crack detection and classification method based on a genetic algorithm (GA) for optimizing image processing technique parameters (IPTs). Under various complex photometric conditions, the crack detection results of concrete infrastructure surface images remain noise pixels. Following that, a deep convolution neural network (CNN) method is used to automatically classify crack candidates and non-crack candidates. Furthermore, the proposed method is compared to state-of-the-art crack detection methods. The experimental results validate the reasonable accuracy in practice. The final goal was to create a crack map, which necessitated automatic pixel-level accuracy.

Author(s) Details

Cuong Nguyen Kim
Faculty of Highway & Bridge, Mien Trung of Civil Engineering, Vietnam.

Kei Kawamura
Graduate School of Science & Technology for Innovation, Yamaguchi University, Japan.

Hideaki Nakamura
Graduate School of Science & Technology for Innovation, Yamaguchi University, Japan.

Amir Tarighat
Department of Civil Engineering, Shahid Rajaee Teacher Training University, Iran.

View Book :- https://stm.bookpi.org/CASTR-V3/article/view/1411

Vision-based Measurement of Geometric Parameters of Cracks in Concrete

The geometrical parameters of a cracked concrete surface are estimated using an optical microscope and an 8-bit RGB image generated with a high resolution camera based on close distance photography. The image’s pixel intensity distribution can be used to determine factors such as fracture breadth, depth, and shape. To estimate the crack’s geometrical dimensions, the image is converted to 16-bit grey scale, and then a mathematical relationship connecting the intensity distribution to the depth and width is derived using the enhanced image. For the crack samples utilised in the research, this connection allows for a 10% and 15% accuracy in estimating the width and depth, respectively. For mathematical processing of picture data and statistical computations of geometric parameters of concrete cracks, OriginLab tools were employed. If the 8-bit RGB image is synthesised from photos of fractures collected with different light directions, the accuracy should be increased even more.

Author (s) Details

Yuriy Vashpanov
Department of Physics, Odessa State Academy of Civil Engineering and Architecture, Odessa, 65029, Ukraine and Public Safety Research Institute, Konyang University, Nonsan, Chungnam, 32992, Republic of Korea.

Jung-Young Son
Public Safety Research Institute, Konyang University, Nonsan, Chungnam, 32992, Republic of Korea.

Gwanghee Heo
Public Safety Research Institute, Konyang University, Nonsan, Chungnam, 32992, Republic of Korea.

Tatyana Podousova
Applied Mathematics Department, Odessa State Academy of Civil Engineering and Architecture, Odessa, 65029, Ukraine.

Yong Suk Kim
Public Safety Research Institute, Konyang University, Nonsan, Chungnam, 32992, Republic of Korea.

View Book :- https://stm.bookpi.org/AAER-V11/article/view/1248