Asphalt pavement surface repair area detection based on smartphone sensors

Abstract

Asphalt pavement repair areas affect pavement performance and service levels. It is necessary to distinguish the repair areas from normal sections. Based on vehicle vibration signals, this study identified ten pavement repair areas and divided them into four cases by factors including length and form in conjunction with the driving approach. Additionally, time domain analysis, frequency analysis, and probability distribution analysis were used to form the characteristics of the repair cases as well as the normal sections. It was found that the maximum value, extreme deviation, standard deviation in the time domain, maximum amplitude in the frequency domain, and peak of the probability density curve would serve as judgment indexes. A framework for identifying the repair areas was also established based on the five indexes. By validation, the overall accuracy can reach 95.0%, demonstrating a strong generalization capability.