The Research And Implementation Of Lane Departure Warning System Based On DM6446

With the global development of urbanization and extensive use of vehicles, transportation problem gets more and more people’s. Based on analysis of the cause of a large number of fatal accidents lane departure accidents take up a large proportion which dues to driver’s fatigue, inattention, and other subjective factors.Accordingly, it is necessary to provide a system to alert the driver before lane departure happened, it can assist the driver to keep vehicles in correct driveway, reduce traffic accidents and enhance vehicle’s active safety efficiently.Lane Departure Warning System(LDWS)is one of the popular research field of the intelligent driver assistance systems, generally referring access to the road front information by visual sensor, combined with vehicle speed and other related parameters of vehicle moving state and the set of warning time from the driver, to determine whether there is a trend of lane departure from the vehicle driving lane, if system judge the vehicle be in departure state, through the way of sound or vibration issue the lane departure warning to the driver, assist the driver to avoid or reduce the lane departure accidents.The paper introduces the domestic and foreign lane departure warning method and puts forward lane departure warning system based on DM6446DSP development board.In this paper, it is divided to three modules to state:the image pre-processing technology, lane departure warning algorithm and the software and hardware architecture of system.In the image pre-processing technology module, it describes characteristics of the traditional median filter and puts forward the rapid median filter to the image noise suppression in ensuring that inhibit the image noise, while reducing the time complexity of the algorithm. Subsequently, the paper presents the adaptive equalization algorithm for image enhancement. In the edge detection, the traditional Sobel edge detection has been improved, and using the optimal threshold Sobel operator to extract edge information.In the lane departure warning algorithm module, it describes the Hough transform techniques including giving the analysis of advantages and disadvantages of the traditional Hough transform, improving the lack of traditional Hough and reducing the overall processing time complexity. it uses the least squares fitting method to fit the lane line equation in the image coordinate system. Subsequently, while the vehicle is in motion, when the position of the vehicle change in the road, CCD camera to capture images in the sequence lane characteristics also change the angle, angle between the proposed method uses to determine whether vehicle has deviated from the lane line.In the system of the hardware and software architecture module, it introduced the structure of the DaVinci technology, including the structure of the processor and software platform framework. This paper uses DM6446processor as video processing unit, the ARM and DSP dual-core features that make its application flexible and good for video data processing, but also make development more difficult, this paper used the software framework that TI provided, proposed the use of Codec Engine framework to build software platform, finally compiled to be the executable Server binary file.The last part of the thesis introduces the realization process of this system, include:first, building algorithm testing platform; second, optimize the efficiency of algorithm; third, transplantation algorithm to the MontaVinsta Linux platform. Finally, this paper uses a real traffic scene video to test the system’s efficiency and accuracy by experiments. And the results show that the system has a good effect on vehicle detecting.