A Survey of Vision-Based Road Parameter Estimating Methods.

Intelligent vehicles need to acquire real-time information on the road through sensors, calculate the limit of car speed and angular speed, so as to provide safety for the control decision. We argue that the road conditions such as snow, ice or humidity pose a major threat to driving safety. We divide the current methods of estimating the road parameters based on the visual sensor, as friction coefficient estimation method, road curvature estimation method and the road slope estimation method. The significance of various methods to intelligent driving, the current research status, as well as scientific difficulties are discussed in detail. Finally we discuss the possible research directions, including establish large-scale open data set, road status prediction methods under multi-task constraints and online learning mechanisms.