Research On Adaptive Threshold Lane Detection Methods Based On LiDAR

As one component of the driverless environment perception,lane marking detection is highly valued due to its direct influence on the safety performance of autonomous vehicles.The commonly used method is the camera-based image processing technology,which is relatively mature,but still cannot overcome the impact of lighting conditions,especially at night.The Li DAR collects high precision and three-dimensional environmental information without being affected by the ambient light,makes the detection method based on Li DAR become the focus of research in recent years.Nevertheless,state-of-the-art detection methods mainly rely on the 32 or 64-beam Li DAR,resulting in extensive amount of data that makes it impossible to guarantee real-time performance.By considering time consumption and cost control,we proposed an adaptive threshold lane marking detection method based on a 16-beam Li DAR.The main contents are as follows:(1)Summarized and analyzed the current Li DAR-based lane marking detection methods.Methods of Li DAR-based pavement point cloud extraction are described in detail.Advantages of methods based on the conversion from three-dimension to two-dimension and the direct processing to three-dimensional data are pointed out,respectively.The main problem is the existence of curb points.Then two kinds of lane marking detection methods based on Li DAR are introduced,namely,echo pulse intensity based methods and pulse width based methods,and the existing problem is low efficiency.(2)Aiming at the problem of curb points,a curb filtering algorithm based on segment point density is proposed.A random sample consensus algorithm with constraints is used to extract the original road data,and the background data is eliminated after analyzing the spatial distribution characteristics of road.Then road data is clustered to different scanlines.By analyzing the distribution statistical characteristics of the roadside points in each scanline in a specific direction,a segment point density method is proposed,and the curb data is filtered out to further optimize results of road data extraction.(3)An adaptive threshold lane marking detection algorithm is proposed to solve the low-efficiency problem.It is well known that lane marking echo intensities are higher than intensities of road data,and the echo intensity of the lane markings decreases with the increase of the distance.Therefore,the Otsu algorithm is improved by redefining the threshold selection interval.The value of the threshold in each scanline is determined by echo intensities of points in different scanlines.It is not affected by echo intensities in the front and back scanlines,which reflects the adaptive characteristics of threshold in each scanline.This method improves the efficiency and accuracy of threshold selection.On this basis,the 16-beam Li DAR is utilized to collect five datasets from different scenes,and the adaptive threshold lane marking detection method is verified.Results show that this method is capable of identifying lane markings efficiently and accurately.