The Key Technology Of Front Vehicle Safety Warning System Based On DM6446

Through a variety of sensors mounted on the vehicle, the front vehicle safetywarning system can provide early warning for the driver and keep the vehiclerunning in right lane and keep a safe distance from front vehicle, and control it underthe specific conditions.Lane detection, detection and tracking of the front vehicle are the keytechnology to the front vehicle safety warning system. To overcome theshortcomings or limitations of lane detecting, detection and tracking for front vehicleincluding:1) the lane model are classified as linear model;2) Hough transform usedto detect lane, is not ideal and conducive to the embedded platform;3) shadowdetection robust features based on the front vehicle shape is short of robustness,intergrated with the embedded platform DM6446,this paper implement front vehicleusing the dual threshold segmentation lane, extract the lane feature points, and themethod of least squares fitting based on two curve the model of the lane, while,using SIFT feature in image, symmetric coding and matching, using Houghtransform to detect the symmetry axis. In the experimental evaluation, the proposedfront vehicle safety warning system has robustness and effectiveness. The maininnovation of this paper is listed followed.Chapter3proposes algorithm of lane detection based on two times curve model.The algorithm in the set region of interest, the integration of access lane candidateand region growing method to set optimal threshold, with clustering method toobtain the candidate points to extract lane candidate clustering center set, followedby further screening and classification of the candidate points to obtain the effectivefeature points, finally detect the lane by the least squares fitting with the effectivefeature points.Chapter4proposes the front vehicle detection algorithm based on optimal SIFTwith symmetry. This model extracts and encodes SIFT of the image in the set regionof interest, while the improved matching algorithm is used to match the sourceimage and the symmetric image matching points, and confirms the center axis ofsymmetry, finally the front vehicle is implemented by Hough.Finally, the hardware performance and various optimization techniques based on TMS320DM6446processor platform is applied to optimize the algorithm:embedded C code, Ping-Pong double buffer, efficient image library and compileroptimization options.The experiment shows that the lane detection and tracking algorithm on frontvehicles presented in this paper have a good performance, and meet the requirementsof the front vehicle safety warning system.