Lane detection with Kalman filter

Lane detection has received much attention, since it is the essential technology for vision-guided vehicle navigation. This paper presents a robust method for detecting lane marks, which is real-time applicable in the field. In general, robust approaches rely on intensive analysis of image entities, which in turn requires expensive computational cost. To resolve this knotty problem, we suggest to utilize the Kalman filter. Our method employs the Kalman filter in such a way that its state parameters may induce expected areas for lanes in the next frame. We then apply a decision theoretic approach for detecting lanes in the expected areas. We conclude the paper with some experimental results that illustrate operational characteristics of our method.