A Study Of Lane Detection Method Based On DM642

With the increasing number of cars, driving safety and simplicity of operation are concerned more in the world, The development of the vehicle safety driver assistance system is becoming a hot spot of vehicle engineering. Lane departure warning system plays an important part in the safety assistant system, and rapid and accurate detection on the lane lines becomes a crucial premise of lane departure warning. This paper studied the lane line detection methods.The road images captured by the car camera contain a lot of useless information to affect the processing speed. In order to accurately detect the lane line, image pre-processing is necessary. Image pre-processing needs some steps such as dividing the region of interest, gray-scale transformation, noise filtering, contrast enhancement, edge enhancement and threshold segmentation. This article uses steerable filter for edge enhancement. First of all, uses edge distribution function method to get the direction angles of the right and left lane lines, which are the input angles of the steerable filter for filtering in a certain direction. In consequence, the direction of the lane line features is made better use to greatly weaken the useless information and the lane is highlight.A combination method of Hough transform and least squares is employed to fit the lane line. For the first frame image, the robust Hough transform is used firstly and then the direction angle of the right and left lane lines supplied by steerable filter is adopted to constraint the polar angle of parameter coordinates, leading to a great reduction in the computation of Hough transform. Secondly, the Kalman filter algorithm is applied to lane tracking, and within the tracking area,the use least squares method is to fit the lane line. This will not only ensure maximum fitting accuracy but also can meet the real-time requirement.In this paper, the core processor of the hardware system is TMS320DM642. In order to shorten the development cycle of the algorithm, early the software simulation is done in VC++ platform, and then the algorithms are written in C language in DSP, and optimization of the algorithm. Finally a hardware experiment is done on DM642 board.