Study On Strategy Of Lane Departure Warning System Based On Machine Vision

At present,traffic injuries have become one of the ten factors of human’s death.Among these injuries,nearly 50% of traffic accidents are caused by vehicles deviating the current lane,so aiming at the lane departure while driving there is important practical value developing a Lane Departure Warning System.At present,with the rapidly development of machine vision,the technique of lane detection is gradually maturing,and developing reasonable and useful warning strategy based on related detecting information has become the key of this research field.Based on this,after investigating a large number of documents,this paper proposes a warning strategy that the main parameters of the strategy are vehicle’s navigation Angle and the horizontal distance.The algorithm can avoid complex camera calibration,improve detection speed while reducing the complexity of system,validate the proposed method basing on data of road experiment.The main research of this paper includes:(1)Preprocessing lane image.In order to reduce the influence of environmental noise and improve the detection results of lane markings,firstly,carrying out the contrast experiments about collecting lane image by graying,filtering and threshold segmenting,eventually,choosing the image preprocessing methods applied to the research of this paper.(2)Analyzing the actual model of highway and detecting methods.Based on the features of lane departure,this paper chooses straight lane as the main object,detects lane efficiently by improving Hough transform for straight line detection.Based on that,Kalman tracking model is established to track lane;(3)Researching lane departure warning model.After analyzing the three classic lane departure warning model,I propose a method for lane departure warning that camera calibration is not needed and analyze its rationality and research the calculating methods of related parameters;(4)Experimental verification and result analysis.Based on the theoretical research,this paper builds the lane departure warning model in the Simulink software and validates and analyzes the warning strategy based on data of lane image.The experimental result shows that the proposed methods can achieve real-time lane detection,and warn drivers when changing lanes with cornering lamp off or departure from the lane.