Effective Road Lane Detection and Tracking Method Using Line Segment Detector

In this paper, we present an effective lane detection and tracking system using a fusion of Line Segment Detector (LSD) and Kalman filter. We employ line segment as low-level features to detect lane markings on structured road. Firstly, we segment the road surface region through adaptive method, and then a different grayscale method is used to achieve strong contrast than the previous studies. Secondly, we apply LSD algorithm on them and remove incorrect line segments in an adaptive way. Finally, two Kalman filters are used to track the virtual boundary points, which also be applied to predict the next processed region of interests (ROIs). Meanwhile, real-time performance can be better achieved. Experiments are conducted on several open datasets which are collected in real-world scenarios, such as Caltech dataset. The results demonstrate the efficiency and the robustness of our proposed method.