Stone pavement distresses detection. An Automatic procedure based on Deep Learning

Computer Vision (CV) makes it possible to use the information content of a given image or video for several scientific and technical applications. The research aims at utilizing CV and deep learning (DL) techniques for the automatic detection of stone pavement distresses. The proposed method can be applied to Pavement Management System (PMS) applications. Several experiments have been carried out on an urban road in Palermo (Italy) using a test vehicle equipped with a video recording system (VRS). Numerous stone pavement distresses have been detected and measured in real-time. The first results obtained with the proposed procedure prove that the differences between the real and the estimated surfaces of pavement distresses are relatively modest and the errors are attributable above all to the camera oscillations due to the pavement irregularities.