Research Of Lane Detection Method Based On DM6437

With the development of socio-economic and increase of car ownership, safetyof driving and simplicity of operation are concerned in all over the world.LaneDeparting Warning System is one of the key technologies of intelligent vehiclesafety driver assistance systems. This thesis study the monocular vision-based lanedetection algorithm in-depth.The main work of this thesis is as follows:Firstly, The traditional Hough transform achieve better lane detection, but thealgorithm is more complex.In order to improve the real-time,this thesis presents amethod based on improved Hough transform.Limitting the detection area,Settingaside most of the invalid information in the image, It can greatly reduces thecomputational volume.By real-time forecast,This thesis make further detectionlimit to improve real-time algorithm.When test results is less than the settingthreshold, End of the track in forecast region. Re-initialize the detection lane.Theexperimental results show this method can solve time-consuming of the traditionalHough transform,and meet the real-time requirements for vehicle equipment.Secondly,Because of a number of shortcomings for improved Hough algorithm,this thesis put forward a road texture feature-based detection methods. First of allto find the vanishing point of the road, then find the best lane lines position withHough transform by the vanishing point and several points around. Then throughthe forecast, the vanishing point and the lane lines can be narrowed in the limittesting space.It can improve the real-time of this algorithm.The experimentalresults show that this method does not affect the improved Hough algorithm forreal-time, It also can effectively inhibit the lane which is not clear and roadobstructions interference.Finally, Algorithms have ported to the TMS320DM6437hardware platformCombine the TMS320DM6437embedded system,this thesis describes the algorithmoptimization methods. including the reasonable allocation of space algorithm,compiler option to optimize and to write assembly code. Realization of the code theapplication layer, the layer of code and the compilation layer multi-dimensionaloptimization, to improve the real-time. The Experiments results show that theoptimized algorithm has better real-time and robustness.