Research On Detection And Warning Algorithm For Curved Highway Based On Binocular Vision

Curve is a common structure in highway.Due to the short sight distance and complex alignment,numerous of traffic accidents have taken place in curves of highway.Majority of the accidents were caused by errors and fatigue driving of drivers.Therefore,it is significant to detect the curves automatically and warn drivers earlier before turning.This paper designs a detection and warning algorithm for curved highway based on binocular vision.This paper studies the image preprocessing methods according to the edge feature and noise disturbance in digital images of curves.The Region of Interest in image is extracted after histogram equalization based on gray statistical features of each pixels.With the analyzation of filtering methods based on space domain and frequency domain,this paper design a method for suppress the noises of digital images of curves based on median filter algorithm.This paper enhance lane marks' edge by using LoG sharpening filter templates,and get binary image by using the OTSU method.Curves are detected through their lanes,by which curves are judged and feature points are extracted.The effects and limitations of least square method fitting lane marks based on b-spline and Hough transform are analyzed in static images.A progressive Hough transform algorithm is designed in this paper according to prior knowledge of lanes.In order to avoid fall out ratio,this paper presents a lane detection algorithm based on dynamic repository.The judgment method and feature points extraction method are designed based on directions of expanded connected fields.For eliminate the low precision of the imaging planes aiming by lines,this paper designs a method of stereo rectification by relative rotary matrix and translation vector through stereo calibration.For improvement of accuracy on distance measurement,this paper corrects the formulas on distance measurement by relative rotary matrix and translation vector.A warning algorithm for curved highway is designed on the basis of judgment method,feature points extraction method and distance measure method.Performance of the detection and warning algorithm for curved highway is proved in roads and tunnels.The performance of detection method,judgment and feature points extraction method,distance measurement and warning for curved highway are analyzed in this paper.The experimental results suggest that detection and warning algorithm designed in this paper can detect curved highway accurately and warn drivers with good robustness.