Vehicle Identification Based On Lane Departure Early Warning Of Auxiliary Driving System

Advanced driver assistance system is to use sensors to sense the surrounding environment and collect information,and carry out systematic calculation and analysis on it,so as to inform the driver of possible risks in advance,so that the driver can take measures to avoid risks in advance,and increase the safety and comfort of driving.Due to the continuous development of computer vision,advanced driver assistance system based on vision is more economical.Therefore,advanced driver assistance system based on vision has become the focus of researchers at home and abroad.Therefore,this thesis studies the lane line identification,lane departure warning and vehicle identification in front based on vision.(1)Research on lane recognition based on vision.Firstly,the road image is preprocessed.The main step is to determine the ROI;then the weighted average method is used to gray the image to reduce the amount of calculation;then the median filtering method is selected to screen out the noise in the image;the big law is used to identify the contour of the lane line;finally Canny operator extracts the edge of the image.In the traditional Hough transform method,the constraint condition of polar diameter and polar angle is added to realize the accurate location and recognition of bilateral lane lines,which makes up for the deficiency that the original Hough transform algorithm can only recognize unilateral lane lines;and the lane line recognition is realized by the least square method,which requires to limit the search range of lane line feature points,and the search index includes gradient information and lane line geometry information.The results show that the lane recognition algorithm can recognize the lane well.(2)Research on lane departure warning model based on lane lines.Firstly,the existing lane departure warning models are analyzed in detail,mainly including the structure and application conditions of CCP model(Car’s current position),FOD model(Future offset difference),TLC model(Time to line crossing)and KBIRS model(Knowledge based interpretation of road scenes).For the identified lane lines,this paper builds an early warning model based on lane deviation rate,and realizes the early warning judgment of vehicle left and right deviation by setting the threshold range of control quantity.The results of multi-frame image testing show that the proposed method based on lane deviation rate can accurately judge whether lane deviation is a simple,effective and feasible departure warning model.(3)Detection and recognition of vehicles in front of the lane by the auxiliary driving system.It mainly includes the determination of the area where the vehicle may exist and the vehicle identification based on RBF neural network.Firstly,a series of morphological operations are carried out on the shadow features of the vehicle underbody to determine the possible hypothesized regions.Due to the shade of trees and other shadows,the system may misjudge,so it is necessary to establish an recognizer to further judge whether the hypothesized area is a car.By extracting 19 characteristic parameters,such as region description and shape description,the RBF neural network model was built to further confirm the presence or absence of vehicles in the determined hypothesis region.Finally,the positive and negative sample libraries of vehicle images are established,and the RBF neural network recognizer is trained until it converges.The validation of random test samples shows that the neural network recognizer constructed in this paper can reliably identify vehicles in the search area,and the accuracy rate reaches 94%.