Extraction Of Overhead Contact System Based On Mobile Laser Scanning Point Cloud

The overhead contact system plays an important role in providing power for electric locomotives and is a major component of the railway system.The extraction of overhead contact system can provide important data support for masts detection items including appearance,side boundary,tilt angle,and angle with center line,as well as contact wire detection items including conductor height and stagger,which is of great significance for ensuring the normal operation of overhead contact system and maintaining railway traffic safety.At present,the extraction accuracy of the masts is not high due to the influence of other components,and the extraction accuracy of the contact wire is greatly affected by the slope and length.Aiming at the problem that the extraction accuracy of masts is not high due to the connection with various overhead contact system components such as cantilevers,tie rods,messenger wires,dropper wires,and contact wire,the point cloud extraction algorithm for masts based on convex hull is proposed.The two characteristics of the masts are mainly used: 1)The masts are connected to the cantilevers,tie rods,messenger wires,dropper wires,and contact wire and other components in the horizontal direction with a larger span and higher than other non-ground objects;2)The masts has the similarity above and below.First,under-segmentation is performed.The masts connected to other overhead contact system components are distinguished from other non-ground features by the two attributes of horizontal distance and height.Then the over-segmentation is performed.The main part of the masts is selected to distinguish them from other components of the overhead contact system.Finally,the completed masts are extracted in the vertical direction by the convex hull generated by the horizontal projection of the main part of the masts.This paper verifies the effectiveness of the proposed algorithm by extracting masts from the vehicle-borne 3D laser point cloud data of 1.285 km long station.The vehicle-borne 3D laser point cloud data of including a set of 51 masts and their connected cantilever,tie rods,and other components is compared with a point cloud extraction algorithm for masts based on convex hull and the point cloud extraction algorithm for masts based on spherical neighborhood to compare the accuracy of the two algorithms.The Precision of the point cloud extraction algorithm result for masts based on convex hull reaches 100.00%,the Recall reaches 98.04% and the F-Measure reaches 98.99% at the object level;the Precision,Recall,and F-Measure has reached 99.01%,94.64% and 96.77%respectively at the point cloud level,compared with the point cloud extraction algorithm for masts based on spherical neighborhood,the Precision is equivalent,the Recall is increased by27.42%,and the F-Measure is increased by 16.37%.The extraction of railway contact wire using high-Precision Li DAR point cloud data obtained by 3D laser scanning technology is of great significance for ensuring railway traffic safety.Aiming at the difficulty of applying the existing method to railway scene contact wire extraction due to the influence of slope and length,the point cloud extraction algorithm for railway contact wire based on the height difference of the nearest rail points was proposed.The height difference between the railway contact wire and the rail plane is relatively stable.The height difference between the candidate contact wire point and the nearest rail point is calculated,and the height difference range is determined by analyzing the vertical distribution of components of the candidate contact wire point cloud and rail to complete the extraction of the contact wire.By selecting two groups of railway point cloud data with a length of about 500 m,a slope of 0.2‰,1.0‰,2.0‰,2.8‰ and 3.8‰,and a slope of about 4.0‰,with a length of 200 m,400m,600 m,800m and 1000 m,the contact wire extraction experiments were performed using the point cloud extraction algorithm for railway contact wire based on the height difference of the nearest rail points and the point cloud extraction algorithm for contact wire based on elevation distribution.The Precision,Recall and F-Measure of the result of the point cloud extraction algorithm for railway contact wire based on the height difference of the nearest rail points are 24.03%、53.44%and 45.75% higher than the point cloud extraction algorithm of contact wire based on the elevation distribution at the point cloud level,and the accuracy indicators have been maintained at more than 93%.The results show that the point cloud extraction algorithm for railway contact wire based on the height difference of the nearest rail points is less affected by the slope and length of the terrain,and can be applied to the cloud extraction of the contact wire point cloud in the railway scene.