Study On Lane Shift Early Warning Lane Detection Technology

With the development of society and the improvement of people's living standard,cars has become more and more popular.It has become an indispensable means of transportation in people's lives.At the same time,the problem of traffic safety has become more and more severe.In order to improve vehicle safety and reduce traffic accidents,lane departure warning system arises at the historic moment.There are many mature lane departure warning systems,but they are costly and expensive.The traffic recorder is relatively cheap,and now the function of the recorder is single.Therefore,developing a traffic recorder with lane departure warning function is of great significance and economic value.According to the lane detection algorithm in the existing Lane shift early warning system,the related algorithms,edge detection algorithms,lane line recognition and tracking algorithms,and the model of lane migration early warning system are studied.In the process of lane grayscale processing,compared the weighted mean value method,the mean value method and the maximum value method,the weighted mean value method is adopted,it is more applicable.In the process of image smoothing,spatial domain method and frequency domain algorithm are analyzed.Through experimental simulation,a median filtering method with good speed and good results is selected.Finally,the image two valued algorithm is studied,and the iterative threshold method and Otsu threshold method are compared.It is found that the Otsu threshold method is more adaptable.In the analysis and contrast of lane detection algorithm,the classical edge detection operator and the edge detection algorithm for mathematical morphology are analyzed.On this basis,a mathematical morphological edge detection operator based on multi angle and multi scale is proposed.The algorithm proposed in this paper has a good detection effect,but its running time is long,and it does not meet the requirements of real time,because the amount of information entropy processing is too large.In order to improve the algorithm,we set a fixed weighting coefficient instead of the weighting coefficient obtained by information entropy.Through statistical information entropy weighting coefficient,it is found that these coefficients tend to a fixed value.After comparing the running time of the improved mathematical edge detection operator and the classical edge detection operator,the operation time of the improved mathematical morphology edge detection operator costs the least time,and the speed is about 3 times higher than that of the classical edge detection operator.Finally,the improved mathematical morphology edge detection operator is adopted as the edge detection operator in this paper.This operator can not only meet the real-time performance,but also improve the robustness of the system.In the analysis and contrast of lane recognition methods,the Hof transform of two valued images is used to find a peak point,and then the intersection point of the corresponding line and the horizontal axis of the image is used as the hidden point of the image.The region is dynamically divided by the polar angle constraint of the hidden point,and the region is more adaptable than the fixed sense region..Then,the Zhang-Suen fast thinning algorithm and mathematical morphology skeletonization deburring algorithm are used to verify the effectiveness of the algorithm.Then,Hof transform straight line fitting algorithm and least square parabola algorithm are used to fit the straight lane and curve Lane respectively,and the fitting effect is good.The lane line recognition of the samples of different scenes can be detected in general,and the detection rate of the experimental samples collected in this paper is 93%.In order to further improve the recognition rate and accuracy of lane detection,this paper uses Calman tracking matching model to track lane lines.Through simulation,the detection rate of samples is 95.6%,the average time of handling each frame is 0.06 s,and the standard and real-time performance of lane detection is achieved.In the study of lane migration early warning system model,the commonly used lane departure warning system is compared and studied.In this paper,the lateral offset distance and relative navigable angle are derived by using the pole and pole diameter of the detected lane lines as the parameters,and an early warning model of lane lateral migration based on the lateral offset distance is proposed.And an early warning model based on relative deviation angle.The feasibility of the model is verified by simulation.