Robust detection system of illegal lane changes based on tracking of feature points

This study proposes a robust real-time system to detect vehicles that change lanes illegally based on tracking feature points. The algorithm in the system does not need to switch depending on the illumination conditions, such as day and night. The camera is assumed to be heading in the opposite direction to the traffic flow. Before starting, the system manager should initially designate several regions that are utilised for detection. Then, the proposed algorithm consists of three stages, such as extracting feature points of corners, tracking the feature points attached to vehicles and detecting a vehicle that violates legal lane changes. For the feature extraction stage, the authors used a robust and fast algorithm that can provide stable corners without distinguishing between day and night or weather conditions. Salient points are selected among the corner points for registration and tracking. Normalised cross-correlation is used to track the registered feature points. Finally, illegal change-of-lane is determined by the information obtained from the tracked corners without grouping them for segmentation. The proposed system showed excellent performance in terms of the accuracy and the computation speed.