Research On Highway Lane Detection Method

With the increase in the number of vehicles, the number of car accident victims has increased annually. Many accidents are caused by a lack of awareness about driving conditions due to driver carelessness or visual interference. Lane detection based on vision technology is the key technology of safety driving systems. The lane is the most important information in driving environmental. Lane detection for intelligent vehicle in highway environment has been focused on this dissertation. Lane feature extraction and lane model fitting problem under different light conditions have been researched in this dissertation. This thesis mainly includes the following three aspects:Through the analysis of image gradient and combining with the actual image, this paper proposes a feature extraction method based on gradient-enhancement. LDA calculates the gradient enhanced transform vector which produces a new gray-level image from an RGB color image. Lane feature image is obtained by adaptive canny edge detection. To deal with illumination changes, the feature extraction vector is dynamically updated. Experiments show that under the condition of the illumination change, the proposed feature extraction method can retain the main characteristic of lanes, and improve the validity of the lane detection.A lane detection method of Statistical Hough transform based on gradient constraint is proposed on improving the classical Hough transform grid quantization precision as well as reducing the large amount of calculation. Statistical Hough transform uses the Gaussian kernel function to model each pixel in the image solving the problem of the standard Hough transform grid quantization accuracy. Using the method of gradient threshold limits the size of initial data set for reducing the amount of calculation and improving the follow-up detecting accuracy. Lane parameters of continuous probability density function are given. The lane detection result is used to initialize the particle filter tracking algorithm. The measurement model of particle filter is set as the probability distribution function of Statistical Hough transform. The results of the experimentation show that under highway circumstance the provided method can rapidly and robustly detect the lane.This paper gives an evaluation method in combination with ground truth information to evaluate the performance of lane detection system. This method converts the original video sequence images into TS image using time slice. Ground truth information is obtained by sampling point interpolation operation. And lane detection accuracy degree of error function is given. At last a unified test database access method is presented.