Research On Power Line Detection Algorithm Based On Vehicle LiDAR In The Assessment Of Large Vehicle Fleet

There are many problems in super large vehicles driving,such as small vision,large blind area for drivers,large turning radius and so on.When fleet of large vehicle passes through unknown road conditions,it is difficult for drivers to accurately observation the surrounding environment,which is often easy to cause accidents.In case of impassable road conditions,it will be very difficult and time-consuming for the whole vehicle formation to change direction.The intelligent pilot vehicle loading environment sensing system is an effective solution.The intelligent sensing pilot vehicle automatically detects the road conditions in front,and makes whether it can pass or not according to the vehicle parameters(width,height,turning radius,etc.)of the rear fleet,so as to report to the rear fleet,ensuring the safety and driving speed of the fleet.The power line is an important national infrastructure.Because the power line is suspended in the air without an object of reference,it is difficult for the driver to visually identify whether its height is passable to go through.In order to avoid the damage of power line in the process of driving,it is necessary to detect the power line in real time.Therefore,this paper studies the power line detection algorithm based on vehicle borne lidar.The research and work mainly include:(1)Limitation of horizontal installation of lidar to measure power line was analyzed,and a data acquisition scheme of vertical installation of lidar is proposed.The effective range of some kind of radar is tested.The point cloud dataset of power line are set up.(2)A filtering method based on normal direction is proposed.Tree kind of evaluation index are proposed to evaluate the result of power line point filtering.Experiments show that the filtering method based on the normal direction can retain the power line point cloud,remove most of the other target laser point,and greatly improve the proportion of power line laser feet in the point cloud data.(3)A Euclidean clustering method with normal constraint is proposed,which is improved on the traditional Euclidean clustering method.Through the comparative experiments of multi scene point cloud data,it is proved that the improved method can segment the power line points and other points better than the traditional Euclidean clustering method,especially when the power line points are close to other target point clouds.(4)The process of power line detection in clustering set is proposed,which is mainly divided into three parts: clustering posterior,RANSAC line fitting and power line posterior.A posteriori algorithm of power line is designed to determine which subsets of lines in the line set constitute the power line corridor based on the criteria of scanning line sequence,direction relationship between lines in the line set and the overall direction of the line set.It’s proved that the algorithm can correctly output the power line point cloud.(5)Based on the above methods of filtering,clustering and power line detection in clustering,a complete set of power line detection algorithm based on vehicle lidar is proposed.Firstly,the original data is preprocessed by direct filtering to remove the laser points of low elevation and radar carrier;secondly;secondly,the normal direction filtering is carried out to retain the power line point cloud and remove most other points;then,the normal constrained Euclidean clustering is carried out to obtain a series of clustering point sets;finally,the power line detection process is carried out in each clustering set to output the point cloud of correct power line.Experiments show that this method can detect the power line point cloud data of various types of radars and scenes in real time.It has high precision rate and recall rate from the laser point level,and low false detection rate and missing detection rate from the example level.It can be used as a part of the intelligent pilot vehicle environment perception algorithm in the project of large motorcade trafficability detection.