A Machine Learning Approach for Pavement Crack Detection and Classification

Abstract

Pavement cracking, a common issue affecting road infrastructure, significantly impacts road performance and longevity. This article employs MATLAB software to process pavement crack image detection, aiming at a machine learning approach for pavement crack detection and classification. A comprehensive model is utilized for categorizing various cracks and evaluating detection confidence. Additionally, a dedicated crack segmentation network is employed to achieve precise pavement crack segmentation. This approach incorporates advancements that improve precision in crack classification and segmentation. Based on the segmentation results, computations were performed to determine the length of linear cracks and the area of alligator cracks. Research findings demonstrated exceptional accuracy in recognizing block cracks, alligator cracks, transverse cracks, and longitudinal cracks. Notably, longitudinal and transverse cracks exhibited high detection rates, while alligator and block cracks have lower detection rates.