Real-time lane and vehicle detection based on a single camera model

This paper presents a real-time lane detection algorithm and a fuzzy-based vehicle detection approach for autonomous vehicles. The fast adaptive lane detection algorithm is efficient and robust during the day and at night. Modified global lane models are proposed to increase the accuracy of the detection. Moreover, an automatic adjustment method is presented to adapt to the changes of lane width and the inclination of roads in different situations. The developed fuzzy-based vehicle detection method, contour size similarity (CSS), compares the projective sizes of vehicles with the estimated ones by fuzzy logic. Vehicle detection aims to detect the closest vehicle preceding the autonomous car in the same lane. Results of the experiments demonstrate that the proposed approach is effective in distance estimation and both lane and vehicle detection. Furthermore, the approach rapidly adjusts to changes of detection target when another car cuts in the lane of the autonomous vehicle. The obtained information supports the automatic driving of autonomous vehicles. This integrated algorithm of lane and vehicle detection was verified on the autonomous car, TAIWAN iTS-1, onexpress highways and in urban environments at velocities of 110 and 50 km/h, respectively.