Traffic Lane Line Classification System by Real-time Image Processing

The traffic safety has been a major concern in recent years. One of the effective approaches to prevent the traffic accident is to develop advanced driver assistance systems which can alarm driver in dangerous situation. In fact, changing lane or overtaking another vehicle is one of the most dangerous driving behaviors. Therefore, it is important for drivers to recognize current lane line types to take proper actions. However, classification systems proposed so far can only distinguish up to five types of lane lines, such as dashed and solid. Hence, the existing road classification systems are not suitable if there are more types of lane lines on the road. In this paper, an improved method is proposed to classify more lane line types by real-time image processing. In order to increase the detection accuracy of lane line types, the image stitching method is applied to reduce the misjudgment caused by blocked lane lines. A set of features about pixel distribution is utilized in the classifier to distinguish more than five lane line types. Furthermore, the results of experiments which are carried out in real road driving show high accuracy of the proposed classification method under the various situations.