Research On Lane Departure Warning System Based On Surround View Assistance System

Lane departure warning system is one part of safe driving assistance system, at present, domesticand foreign research of lane departure warning system is mainly based on the forward-looking systemof vehicle monocular vision. In this paper, Lane departure warning system based on surround viewdriving assistance system was studied, the main results are as follows:According to the structure of surround view driving assistance system and features of thecollected images, one kind of registration method was designed based on the features of the image,then the calibration of system was completed fleetly, forming a360degrees rendering of the vehicle.The image pre-processing and lane maker detection method were studied. The weighted averageimage pre-processing method was used to gray the image first, and then to median filtering; Severalkinds of edge detection operator’s principle were analyzed, and all kinds of operators were used forsimulation experiments. Finally by comparing the experimental results and considering the followingprocessing requirements, Canny operator which has the better detection effect was chose at last.Through a detailed analysis of the road model, and by taking the characteristics of collectedimages into consideration, a model based on the straight road was selected. Research on the Houghtransform was conducted and it was used to identify the lane mark. The experimental results showedthat the correct identification rate of this method was over95%A Kalman filter group was designed to be used as the lane mark tracking algorithm which couldtrack multiple lane lines simultaneously. Experimental results showed that this algorithm couldgenerate optimal output value by going through Kalman filter group and finally improved thereliability of the lane departure warning system.Several common lane departure warning models were studied in this paper, and a lane departurewarning model based on transverse position of the vehicle was established. The experimental resultsdemonstrated that the proposed model could keep the lane departure warning system’s accurateness.Using a BYD Auto as the experiment platform, lane departure warning system was tested on thereal road. The experiment results indicated that the algorithms which were used by the system hadgood timeliness and robustness, and the lane mark line correct detection rate of the system reached98%, while warning accuracy rate above95%.