Deep Learning-based Pavement Cracks Detection via Wireless Visible Light Camera-based Network

Abstract

Road maintenance is important to ensure the normal usage of road transportation. Due to complicated effects including natural consumptions and human overuses, various forms of damages appear on the road surface, and among them, pavement cracks are the most representative ones. Therefore, the automatic detection of pavement cracks is important and essential in road maintenance. In this study, deep learning techniques are incorporated to fulfill the automatic pavement cracks detection based on a wireless visible light camera-based network. Technically, a visible light high-definition camera is equipped at the bottom of a maintenance vehicle, and pavement images are automatically captured along with the movement of the vehicle. Then, images are wirelessly transferred to a workstation on the maintenance vehicle, and the pavement cracks detection task is fulfilled via deep learning techniques from the classification perspective. Experimental results demonstrate that, the above wireless visible light camera-based system is capable to discern pavement cracks accurately. Also, real-time performance is guaranteed within this deep learning-based pavement cracks detection study.