Research On Detection And Recognition Of Road Markings Based On Lanes

Driverless vehicle is one of the important research fields of intelligent vehicle. In the study of this subject, the perception of the surrounding environment and road information is one important part. Among the road information, there are the lane lines and markings information. Effective access to the right lane and road sign information will play an important role in driverless vehicles’ decisions. The paper proposed effective detection of and recognition of markings based on lanes. The main tasks include:Firstly, the paper uses the horizontal luminance difference method to detect lane. Converting the road image into YUV color space, setting the differential luminance value and the threshold value to obtain points of the lane edge, and using the method of least squares to fit the lane edge points, then lanes are detected.Then, this paper raises two kinds of detection methods for road markings, which are detection method based on lanes and detection method based on the luminance difference. Detection method based on lanes uses the contour information to determine the scope of candidate markers. Detection method based on the luminance difference is to obtain edge points of the lane and marking by the luminance difference processing, cluster the edge points to separate the lane and marking, and fit the marking.Furthermore, the paper also presents two different road sign recognition methods, namely recognition method based on Hu moments and recognition method based on the lane position relations. Recognition method based on Hu moments is to calculate Hu moments and Mahalanobis distances between the detected marking and each type of sample marking databases, and then recognize the detected marking. Recognition method based on the lane position relations is to consider the position relationships between the lane and marking to complete the recognition. In addition, the experimental data shows that the recognition rate of two methods proposed in this paper reached94.5% and87.5% respectively, with the same200testing direction marking pictures.