Applications of Machine Learning and Deep Learning in Pavement Crack Detection and Characterisation: A Comparative Approach

Cracking in asphalt pavement structures is one of the major forms of deterioration that not only jeopardises the traffic safety but also compromises the durability and serviceability of the roads. Traditional crack evaluation techniques involve manual human field surveys, which has major flaws such as, requirement of intensive labour force and time, poor repeatability and reproducibility of the data and data collection method and subjectivity of the rating measured by the individual surveyor. Accurate detection and characterisation of the extent of cracking is crucial for the maintenance and longevity of pavement infrastructure. Recently, machine learning (ML) and Deep Learning (DL) methods have arisen as an instrumental tool for crack detection and characterisation. The current study provides a comprehensive analysis of the ML and DL applications in crack identification and characterisation to enable a data-driven pavement management system. A number of machine learning techniques such as; Support Vector Machines (SVM), Artificial Neural Networks (ANN), Convolutional Neural Networks (CNN), Recurrent Neural Networks (RNN) and deep learning (DL) techniques were analysed for their suitability in identification of pavement cracks and their extent. The article delves into crack identification and characterisation by focusing on object detection, classification and segmentation methods. Furthermore, the current study discusses the performance evaluation metrics and provides case studies by highlighting the successful application of ML and DL techniques for crack identification and characterisation.