Identification of asphalt pavement transverse cracking based on 2D reconstruction of vehicle vibration signal and edge detection algorithm

Abstract

Transverse cracking is related to asphalt pavement structural performance. However, the method for transverse cracks identification by vehicle vibration signal has hardly been applied because of an extremely low recognition accuracy. The study aims to precisely identify the pavement transverse cracks by vehicle vibration response. A two-dimensional reconstruction method that transformed the vehicle vibration signal into an image was proposed. Five edge detection operators were used to generate the features of vibration reconstruction images. Besides, this study combined two basic characteristics, Euler number, and total target area. Eight classification models based on the above two features were trained by the RUSBoost integrated classifier. The results showed that the classification model based on the Prewitt operator eight connected domain Euler number and the total target area performed excellently. The testing accuracy can reach 93.5%. The classification model presented in this study has an excellent effect on pavement transverse crack identification and feasibility.