Alert Method Of Lane Departure Based On Fuzzy Clusterings And Fuzzy Pattern Recognition

With the development of intelligent vehicles and safety driving assist technology,lane deviate warning has become the main study contents of the field. Presently, thecurrent evaluation system based on one factor can realize early warning when the vehicledeviate the road, but still has its shortcomings. Under vehicles actual driving conditions,there are complex function relation in the factors that impact lane deviate, inevitably affectthe reliability and validity of the early warning system model. This paper attempted to usefuzzy clustering and fuzzy pattern recognition methods to predict whether the vehicleoccurred lane departure, hoping to improve the accuracy and practicality of the system.Firstly, pre-processed the vehicle road image, using parabolic model to build roadboundary model based on the objective function, achieved lane boundary identification.Extracted several state parameters and index properties based on the analysis of roadstance, according to the basic principle of feature selection, optimized two fuzzy patternrecognition the most representative characteristic parameters.Secondly, expounding conventional method and basic algorithm flow of fuzzy theoryand pattern recognition, selected 15 groups of images as training samples, using thedynamic cluster analysis of the fuzzy equivalent matrix clustering to the sample, classifiedthe vehicle safety degree into several categories, established the standard model library,using fuzzy pattern recognition method based on proximity degree and proximity principleto design the classifier.Finally, proceeded experimental verification, put all groups of vehicle driving data asexperimental sample to the classifier, realize classification identification of the vehicledeviation degree and evaluate the security, verify the quality of the classifier model.The paper was a new attempt which used fuzzy pattern recognition to classify thevehicle lane deviation degree, vehicle road status are different, using fuzzy patternrecognition can classify it perfectly, the average recognition accuracy rate reaches 95%.