Bottom-Up Cracks Detection in Road Pavement in Artificial Neural Network

In recent years, road crack detection is a significant task for detecting surface defects, road safety and road maintenance. But there is lack of approach that able to detect the concealed cracks especially bottom-up cracks, and this is challenging due to poor continuity of cracks, various crack’s widths, and low contrast among cracks and the surrounding pavement. The Artificial Neural Network (ANN) algorithm is proposed for detecting and classifying the structural health status of bottom-up crack in road pavements, which allows detecting surface defects and failures. The approach aims at collection of these signatures by utilizing acoustic-sensors, located at roadside, and also classification of SHS in pavements over the ANN approaches. The various Machine Learning (ML) classifiers such as Multilayer Perceptron (MLP), 2-Dimensional Convolutional Neural Network (2D-CNN), Support Vector Classifier (SVC) and Random Forest Classifier (RFC) are utilized as well as compared for the analysis of crack in road pavement. The experimental results illustrate the probability of associating with great accuracy $92.7 \%$ of MLP, $96.7 \%$ of 2D-CNN, $92.5 \%$ of RFC and ${9 9 . 3 \%}$ of SVC respectively.