Title: Convolution neural network model for an intelligent solution for crack detection in pavement images
Authors: Aaron Rasheed Rababaah; James Wolfer
Addresses: College of Engineering and Applied Sciences, American University of Kuwait, Salmia, Kuwait ' Department of Computer Science and Informatics, Indiana University of South Bend, South Bend, Indiana, USA
Abstract: This paper presents a deep learning solution using convolution neural networks for pavement crack detection. The advancements in machine learning and machine vision open new opportunities for researchers to explore the power of deep learning instead of classical machine learning to solve old and new problems. We propose a convolutional neural network model to detect cracks in pavement. Our solution is based on a multi-layer model that encompasses a raw image input layer, convolutional layers, activation layers, max-pooling layers, a flattening layer and multi-perceptron neural network as classification layers. MATLAB was our development platform to create and test the solution. A total of 500 sample images were collected from publicly-available sources. Sixteen different experiments were conducted to determine the best configuration for the proposed model in terms of the number of features. The results of the experiments suggest that the proposed model is effective with a detection accuracy of 96.6% when correctly configured.
Keywords: deep learning; convolutional neural networks; pavement images; crack classification; machine vision.
DOI: 10.1504/IJCAT.2022.10050313
International Journal of Computer Applications in Technology, 2022 Vol.68 No.4, pp.389 - 396
Received: 11 May 2021
Accepted: 27 Jun 2021
Published online: 01 Sep 2022 *