Research On Road Traffic Marking Algorithm Based On Fusion Of Single Frame Image And Sparse Point Cloud

In recent years,the new generation of science and technology represented by artificial intelligence has developed rapidly,and many new location-based services have placed new demands on traditional navigation electronic maps.GNSS(Global Navigation Satellite System),with the support of ground-based and other enhanced systems,civilian consumer-level real-time positioning accuracy has reached sub-meter to decimeter levels,and a navigation map that matches the accuracy needs to form a high-precision location.service.Especially in the field of transportation,in-vehicle navigation maps need to reflect the spatiotemporal geographic environment at a finer scale to support lane-level traffic violation incident supervision and user driving behavior analysis.As an important link in the development planning of the new generation of artificial intelligence in China,high-precision maps are urgently needed to make up for the blind spots of auto autonomous perception,and to achieve accurate and reliable environmental perception and positioning.It plays a key role in real-time positioning and path planning,and is also an important source and decision basis for autonomous vehicles to obtain real-time dynamic information.As an internationally recognized key to future travel,high-precision maps can provide a large amount of accurate and semantically rich data to help autonomous vehicles understand the surrounding environment at a finer scale,assist decision-making control,and meet a variety of high-level applications in the intelligent era Demand,and it has been recognized as an indispensable condition for achieving autonomous driving at the L3 level and above.In China,the width of the lane line in road traffic marking is generally between 10-20 cm.Even in the case of some rural roads or exclusive dedicated roads,the width of the lane line is only 8cm.To avoid the problem of line crimping during vehicle driving It not only puts forward high requirements on the vehicle’s own "perception system",but also has high requirements on the map data.The road traffic marking serves as a sign to control and guide traffic,as well as the source of the element information and attributes of road geometry,lane geometry and appendages(stop lines,diversion areas,crosswalks,arrows,text,etc.)in high-precision maps.,Plays an important role in high-precision maps,and its precision requirements are even better.Therefore,at present,the extraction of road traffic markings is completed by high-threaded lidar to collect dense point cloud data,but due to the high cost of high-threaded lidar and the huge amount of point cloud data,the production efficiency of high-precision maps is low And updating is complicated and difficult.In view of the above problems,this paper proposes a road traffic marking extraction algorithm based on the fusion of single frame image and sparse point cloud.The algorithm first extracts the edge point and internal point coordinate information of road traffic markings from the image and the point cloud,and uses the calibration parameters of the camera and lidar to register the image and the point cloud.Then,through the mutual verification of the edge points and the internal points to eliminate the false detection data,after removing the false detection data,the road traffic marking edge data is clustered and segmented,and the segmented road traffic marking edge data is geometrically corrected to obtain the final plane.Coordinates and elevation information of edge points are determined by interpolation fitting.Finally,this paper also proposes a classification method for the geometric expression of road traffic markings in high-precision maps,and classifies the extracted road traffic markings according to the classification method and completes its geometric expression in high-precision maps.In this paper,the single-frame point cloud in the KITTI data set is used to simulate sparse point cloud data,and the road traffic marking coordinates extracted by the algorithm in this paper are verified by image-based and point cloud-based methods.It has been verified that the algorithm can accurately and efficiently extract the edges of road traffic markings,and provides a new and reliable solution for the automatic extraction of road traffic markings.