Classification of Earthquake-Induced Asphalt Cracks with a Transfer Learning-Based Hybrid Strategy

Abstract

One of the most popular modes of transportation is the highway. Highways that receive timely maintenance avoid future increases in maintenance expenses. It is crucial to identify highway damage brought on by significant earthquakes. because roadways are used to deliver logistical and humanitarian relief to earthquake-affected areas. Consequently, system applications that automatically identify asphalt deterioration are required. Images of asphalt cracks that appeared in five major Turkish cities following two consecutive severe earthquakes in the Elbistan region were examined in this study. The construction department experts classified these fissures as serious and small. In the following phase, a novel deep learning-based model was used to classify asphalt fractures. In the implementation phase of the proposed model, features were extracted using transfer learning models. These features extracted from different models were combined to create a large feature set. The ReliefF algorithm was used to select the most discriminative features from the extracted features. Popular machine learning algorithms such as SVM, KNN, Naive Bayes, and Decision Trees were used in the classification phase. The best classification results were achieved with the SVM algorithm.