Research And Implementation Of Advanced Driver Assistance System Based On Vision

Due to the rapid development of automobile industry，the increasing vehicles bringconvenience and troubles at the same time traffic accidents happens each year. However，tons ofstudies suggest that most of traffic accidents are caused by drivers’ themselves，Therefore，providingauxiliary driving to relieve the pressure for the drivers is significant. When danger is about to occur，warning the driver in advance will reduce lots of tragedies. Lane Departure Warning System (LDWS)is an important part ofAuxiliary driving system. By calculating the width of the lane and the car’srelative position in the lane，it is designed to warn the driver when it will deviate from the lane line.However，in the LDWS，the detection technology of lane is the key part of the system. Nowadays，openness of theAndroid operating system and its capability of supporting the third-party software，Android phones are increasingly popular in the mobile phone market. Hence，the detectiontechnology based onAndroid system which can be used in the Car Navigation or phone to develop aLDWS application，is useful and convenient for the safety driving assist. Therefore，this papermainly discusses the lane detection based on android platform.From the existing technology level，the main factors affected the reliability of LDWS are theinfluence of weather conditions and the illumination and the bad lane line blocked by the obstacles.This paper employs theAndroid Camera to collect the real-time images and then preprocess theimages.This paper uses lane edge information to detect the lane line，in order to improve the real-time performance of the system，the author gray the images first to remove a large number of color isnot the necessary information. This paper uses median filtering to reduce remove the interferences.then the edge detection this paper uses Canny to detect the lane edges.At last，the author realize theCanny edge detection based on inter-class variance. The experiments prove that this kind of methodenables to get a good edge of lane line under both the good and bad light conditions but withoutrelying on human experience to set the threshold.During the processing of the lane line detection，the linear model of lane line is established tosearch the straight lines by Hough algorithm. Then，the detected lane lines are divided into left orright lane to eliminate the interference of other useless line the influence through narrowing of therange of the lane lines’ slope and restricting the length of lane line. In addition，the vanishing point isused to get the continuous lane line. For the sake of the higher real-time performance of system，theKalman filtering is adopted to locate the dynamic interest area to search the lane lines，according tothe lane lines’ location in the last frame image.Machine learning has been studied as an effective method of detecting the vehicle.Alargenumber of vehicle images in this chapter first obtain positive and negative samples, the choice of Haar-like featureAdaboost learning algorithm used for training, and finally get Cascade cascadeclassifier, the experimental results show the actual road environment, vehicle identification is valid. Toachieve vehicle collision warning section, this chapter for target ranging geometry and vehiclecollision time TTC model were studied.In this paper, the proposed traffic safety warning system on theAndroid platform fordevelopment and testing, the effective implementation of the front of the vehicle lane detection andidentification.Based on the above research, the proposed lane detection and vehicle distance measuring methodhas a certain practicality. Can help the driver in front of the vehicle while driving distance to have amore accurate perception, take corrective action in advance.