Real-time detection and classification of arrow markings using curve-based prototype fitting

Driver assistance systems using situation analysis require a reliable environment perception in order to improve the traffic safety. Many previous works dealt with lane recognition techniques. In this paper we focus on the problem of detecting and classifying supplementary lane markings like painted arrows. We present a general geometric approach using curve-based prototype fitting. Although our method can process data from various sensor sources, we focus on the usage of a monocular gray value camera. Regions of interest are determined based on some preliminary knowledge in order to reduce the search space. Finally, prototypes encoded as arc splines are used for the comparison with the extracted contours of object candidates, which enables both detection and classification. We show that the concept of arc spline models yield a data efficient and real-time capable application.