Intersection detection based on mono-camera sensor

Adrian Gabriel Balanescu, Cristina Laura Sîrbu, Ciprian Orhei

Abstract: In the last decades Advanced Driving Assistance System are rapidly growing and offering new features to drivers in the direction of understanding and modeling the environment with the sole goal of achieving Autonomous Driving. In this paper, we present a novel method for detecting intersection based on a mono camera system using features generated by Hu-moments and a Random Forest (RF) classifier. This pipeline is enriched with information regarding the environment like road landmarks. Our approach obtained decent results in a wide range of scenes that were created to cover all possible scenarios that an auto-vehicle normally can occur on public roads.