17-08-2012, 11:39 AM
Automatic Lane Detection and Navigation using Pattern Matching Mode
Automatic Lane Detection and Navigation.pdf (Size: 977.61 KB / Downloads: 40)
Introduction
Nowadays lane detection systems have been widely deployed in automatic driving systems for outdoor vehicle and driver assistance system. The main purposes are for safety and aid for a driving. There have been various lane detection systems which can mainly be categorized in two groups: using a single camera and using multiple cameras. In the first category, the artificial neural network has been applied for detecting lane. Earlier, most of researches emphasized on using edge detection for lane segmentation from other environments. Also, image transformation for detecting lines were also considered, such as, Hough transform. Later, matching techniques had become more focused. These techniques can be performed by matching salient features to lane features, for examples, the LOIS in [1]. Following that, lane modeling had been introduced and become more popular. The models can be described by 1st and 2nd order polynomial functions [2] and B-spline function [3].
Lane Segmentation
A segmentation of lane from other environments is required in order to classify lane type. This work focuses on examination the image to extract the significant information of the lane images which are the sky, trees (or other constructions) and lanes. This can be accomplished by locating the vanishing line and the line of lane boundary. The details are discussed next.
Vanishing line pre-computation
In order to segment the sky, trees and lanes from an image, the possible location of the vanishing line is computed. Using simple 3-level
Proceedings of the 7th WSEAS International Conference on Signal, Speech and Image Processing, Beijing, China, September 15-17, 2007 44
gray-scale value thresholding, the corresponding area of the sky, trees and lanes in the image can be roughly segmented as shown in Fig. 1. The result of segmentation shows that the vanishing line can be assumed to appear in the same horizontal line of the tree area in the image (see Fig. 2). This is because it is the connected area between the sky and the lanes.
Lane boundary computation
The lane boundary in the image can be used to separate an area of the tree and the lane. It can be considered as a straight line with different slopes. It can also be described using a parabolic equation (by using only first order). This information from the parabolic equation will be used later for lane classification. Consider the linear equation (1) for describing the lane boundary,
Conclusions
The paper presents the automatic lane detection and navigation system. The segmentation and classification of the lane are proposed. The parabola model is deployed for lane modeling and lane type classification with optimal parameters. The proposed system shows desirable performance for lane detection and classification.