The Research Of Lane Detection Methods Based On DM6437

The safety of driving,and the convenience of the manipulation are always important indicators of the driving experience to human being,unmanned driving(or autopilot)based on machine vision,precise navigation,high-precision maps and other sensors is born for this.Besides,unmanned driving will play an tremendous important role in improving road traffic,saving energy and promoting the economic development of the whole society and the world.It is the ultimate goal of the development of the automotive industry.Before the ultimate realization of the driverless system,as an important part of the system,the Advanced Driver Assistance System(ADAS)is the main form of the system at present.The research of lane detection and recognition based on machine vision and pattern recognition is particularly critical.The algorithm and hardware implementation of lane detection and recognition based on monocular vision are studied in this paper.The main contents of research in this paper include:image processing algorithm,the practical application and improvement of the combination of the principle and the hardware.Single channel video acquisition,edge enhancement and segmentation,edge detection and extraction,lane feature extraction and display are based on DSP video image processing platform TMS320DM6437.And focus on the solution of the optimization process and real-time problem on the embedded DSP platform.The implementation process is as follows:When it comes to the section of image processing and algorithm,the coordinate transformation and color space model are listed.Adaptive thresholding method is applied to grayscale image acquired by camera.Then,we get the result of lane edge detection by changing the positive or negative parts of the convolution results of the improved templates in different direction with the pixel points into zeros.Finally,for the current lane line in ROW,it is constrained by the feature of ROW,which in the edge of the left and right lane boundary line,the pole angle parameters in the polar coordinate system are discretized,and then the lane line detection and recognition are carried out with the improved IH algorithm.In lane detection and extraction part,the practical application of the algorithm is elaborated,and some details and overall processes are optimized and improved,making it more suitable for the features of the embedded platform.The video acquisition system is analyzed and researched.Combined with the DM6437 video processing system VSP,the overall arrangement and design of the system is made.The scheme selection is made according to the functional planning of the system.On the section of hardware platform,the software development and design environment applied to lane detection system are explained,the design and implementation of the software program and the processing flow are planned reasonable,the hardware components and processing steps of the system analyzed and designed.The experimental results show that the edge detection method using IS can betterfilter out the interference and get better detection results under different environmental and working conditions.The lane detection method based on IH can improve the accuracy of the detection effect,improve the real-time problems of the system,and reduce the tedious steps.Then the experimental results and analysis of the detection system are given,which shows that the lane detection algorithm has a good effect on the embedded DSP platform.Finally,the methods applied in the article,the places need to be improved of the work are summarized and the future arrangement and development are made..