Night-Time Road Boundary Detection with Infrared Channel Features Classifier

Abstract

Road boundary detection is very important to Intelligent Vehicle (IV) System. Recently, road boundary detection during night-time driving condition attracts more and more attentions. In this paper, we propose a novel and fast method for night-time road boundary detection on infrared images. Firstly, a set of novel Infrared Channel Features (ICF) are proposed for describing infrared image patterns. Furthermore, we proposed an Infrared Edge classifier to generate a task-driven probability edge map. Finally, road boundary extraction is performed on the edge map by two steps: searching available road boundaries and second order polynomial approximation. Experiment show that the proposed method performs well with effectiveness and efficiency.