Cement concrete pavement distress detection based on sensor fusion architecture

This study develops a cement pavement distress detection system based on a parallel fusion architecture. The algorithm of IMU includes two modified roughness indexes. The image subsystem uses the YOLOv7 network to detect four types of distress objects (e.g. longitudinal crack, transverse crack, complex crack, and pothole). The proposed system is evaluated in two types of cement pavement: urban and country. The experimental results validate the average detection rate of the sensor fusion reaches 90%. The proposed fusion strategy can compensate for any single-sensor failure and improve the detection accuracy of YOLOv7.