ROI perspective transform based road marking detection and recognition

Abstract

Road marking detection and recognition play an important role in driving assistance system and intelligent transportation system. We proposed a method for road marking recognition which can be applied to highway and city road based on ROI perspective transform. If we can detect the lane lines on the road, we determine the trapezoid ROI where road marking would show up with the guide of detected lane lines. Then the ROI is transformed to square by inverse perspective transform for accurate road marking detection and recognition. If there are no legible lane lines that we can take use of, we estimate the ROI according to experience and adopt the method of connected domain analysis to extract target. Template matching is used for recognition of markings. Overall, the metrics of the method are as follows. First, the detection of lane lines is based on the ATN model which is fast and robust. Second, the detected lines are used as guide for the search of markings, and thus, there is no need to search the makings in the whole image. Third, even there is no lane lines, for example, the lines is worn out or blocked by the vehicle ahead on the city road, the connected domain analysis makes the method suitable for these situations.