An Automated and Integrated Sensing System for Road Monitoring using UAV Images and an Optimized R-CNN

One of the most relevant, but at the same time most time-consuming and costly, aspects of the infrastructure system is the monitoring of road infrastructures, often subject to deterioration that compromises their use. Current monitoring systems consist of individual reports or the use of human resources that, through equipped vehicles, have the purpose of carrying out a reconnaissance process, which is often characterized by errors and uncertainties. In this context, the aim of this work was to experiment and implement an experimental and innovative Automated and Integrated Sensing System (AISS) for the monitoring of road infrastructures. This system, starting from Remote Sensing images from Unmanned Aerial Vehicles (UAVs), uses a Mask R-CNN neural network to identify road cracks. This information, together with other information, is included in a database, which is then used in a Geographical Information System (GIS) for relative visualization. This work therefore proposes a methodology for the implementation of a system that helps policy makers in determining the most urgent interventions. In fact, a categorization of the severity of degradation and a user-friendly visualization, allow us to make decisions based on data.