Research On Lane Detection And Tracking Algorithms Based On Active Voting

Due to the sharp increase in the number of cars,frequent traffic accidents have become a major problem in the current world.When driving for a long time,the human driver bears heavy load,its physiological and psychological state is easy to fluctuate.They can not maintain concentration.The risk of traffic accidents has increased.In order to improve driving safety and comfort,advanced driver assistance systems based on lane detection and tracking technology have become a research hotspot in recent years.This paper studies the lane detection and tracking algorithm based on active voting method for structured roads.Main tasks as follows:Firstly,the road image is grayed and median filtered.The top view of the lane is obtained by inverse perspective transformation.The region of interest is set to reduce interference,and the histogram equalization is used to enhance the image to reduce the influence of brightness change.Different edge detection operators are analyzed,and the Canny operator is selected for edge detection based on the comparison between the principle and the actual application effect.The Hough transform is used to detect the straight line in the edge map.For the characteristics of the road image,the calculation of the Hough transform is greatly reduced by constraining the polar angle.A lane detection algorithm based on active voting is proposed.The active performance is assumed to be a lane line in each line.The characteristic attribute is used as the lane line standard to vote on other lines,and the lane is selected according to the total number of votes.After the lane is fitted stably,the Kalman filter is used to track the lane line,The region of interest is reduced,which greatly reduces the calculation amount.The real-time performance of the algorithm is improved.The fitted lane is more accurate.In order to shorten the development cycle and develop the lane detection and tracking algorithm quickly,this paper uses the Visual Studio 2010 development platform to complete the algorithm design on the PC side with the rich image processing algorithm in the OpenCV vision library,and the performance of the algorithm was verified using an offline lane line video set.The high-performance TMS320DM642 DSP is selected as the hardware development platform for vehicle visual recognition,and the verified algorithm is embedded in C language.The "camera + DM642+ display" hardware test platform was built to provide steering control parameters for the lane keeping system of the test vehicle.The road test showed that the proposed algorithm has certain robustness and real-time performance,in order to realize the lane keeping of the test vehicle.Features provide reliable visual perception parameters.Figure [41] table [3] reference [65]...