NeuroVI-based snow ruts and drivable area detection on snow-covered road

Abstract

Detection of safe driving areas on the road surface is a necessary prerequisite for path planning in snow-covered scenarios. The snow rut trajectory of snow-covered roads can provide the consultation of safe zones' detection. This study proposed to use a Neuromorphic vision (NeuroVI) sensor to detect snow ruts and predict the safe driving area. We establish an image clustering method based on NeuroVI event points' thickness and make a rut dataset under snow scenes. This dataset not only includes the recognition of optimal safe ruts but also includes adjacent boundary points and remote preview points. Then, a feature fusion model with a detection head is designed to output the recognition results, which can provide the safe zones' position in snow-covered autonomous driving.