Dipartimento di Ingegneria informatica, automatica e gestionale

Each year, car accidents impact billions of people, resulting in numerous casualties. Consequently, road safety remains a top priority for nations worldwide. This project aims to enhance driver safety through a feedback system that relies solely on a monocular camera mounted atop the vehicle. The proposed system is a real-time application designed to warn drivers of imminent road hazards, which are classified by their level of risk. Our method employs various computer vision techniques and incorporates a simple 2D-3D correspondence to estimate the longitudinal and lateral distances of objects ahead of the vehicle, under certain simplifying assumptions. The system conducts a comprehensive danger analysis by evaluating potential hazards within the vehicle’s path. Depending on the danger level, warnings are delivered to the driver with varying degrees of invasiveness, using haptic feedback. The proposed method was tested on the KITTI dataset, yielding positive results.