Vision-based detection and classification of pavement mark using neural network for autonomous driving system

This paper proposes an algorithm for an autonomous driving system which detects a pavement mark in an image of the road in front of a vehicle and identifies the mark. The algorithm uses edge pairing to find a pavement mark then identifies the type using a neural network which uses the horizontal and vertical projection of the founded mark as input. The network successfully classified 1073 of 1088 images. The result can be used to provide the accurate position of the vehicle in in-vehicle navigation systems.