TAF PUBLISHING

Lane detection is key to advanced driver assistance systems to avoid traffic accidents caused by driver’s negligence, thereby improving driver’s safety. However, most lane detection algorithms are prone to error in challenging conditions when maneuvering in high curvature lanes, strong backlighting environment, low contrast night, and heavy rain condition, rendering unreliable and hazardous detection motion. This work proposes a lane detection algorithm by robust binary lane marking identification for lane feature extraction. The algorithm combines the median local threshold, line segment detector, and binary line segment filter to remove the noise generated when operating in the above challenging conditions. After correct lane feature extraction, Hough transformed and optimized random sample consensus parabola fitting are applied to detect lane markings. Experiment results show that the proposed algorithm outperforms the previous work in achieving correct detection rate at 95% in real-time.