Study On Power Line Extraction From Airborne LiDAR Point Cloud And Automatic Detection Of Safety Distance

Electricity provision is vital both to family life and to business activities.Smart grid,an electrical infrastructure characterized by reliability,flexibility,efficiency,sustainability,etc.,was first proposed by The Energy Independence and Security Act of 2007 in the United States,which puts higher demands on the safety maintenance of infrastructures such as electrical grids.Examining the distance between power lines and trees under them is the process to eliminate possible outage caused by short circuits because of tree limbs falling on or in power lines.Personnel inspection is a traditional way but it lacks of efficiency and is a labor intensity hard job.Moreover,field inspection is almost impossible in remote mountainous areas.Airborne LiDAR,as a new type of active sensor that has developed rapidly in recent years,can directly acquire high accuracy discrete 3D point cloud on the Earth’s surface,which can be used for determining the distance between power lines and trees,providing a prospective replacement for personnel inspection.In this paper,the raw airborne laser point cloud data of power line corridor were adopted.Studies include automatic extraction of power line from point cloud in complex natural background and Automatic detection of the safety distance between power lines and trees.The main achievements of this paper are as follows:(1)Accurate extraction of point cloud of power line.In this paper,regional growth combined with SVM is adopted for this purpose.Firstly,11 direct and indirect features of point cloud which are beneficial to power line classification are extracted,and the optimal feature neighborhood extraction radius and SVM classifier parameters are determined through experiments.Using the trained classifier to roughly extract the power line point cloud.Aiming at the debris and error in the initial extraction of power line,a region growing algorithm is proposed to eliminate the small debris region.Experiments show that the proposed method achieves high power line extraction accuracy for data with large terrain differences,Retaining the spatial structure and extensibility of the power line.(2)Relationship between point cloud density and power line extraction accuracy.In this paper,power lines are extracted from 8 sets of power line corridors with different densities.The results are quantitatively analyzed by three indexes:overall test accuracy,error I and error II.The experimental results show that in order to achieve accurate extraction of power line point cloud,the airborne LiDAR point cloud data density should be above 10pts/m~2.This conclusion has certain guiding significance for researchers of airborne LiDAR.(3)Automatic detection of power line safety distance.The specific process includes four steps:power line segmentation,single power line separation,single power line 3D model reconstruction,power line safety distance calculation.Firstly,an algorithm of K-means clustering under large segmentation threshold is proposed to determine the central coordinates of the electric tower and realize power line segmentation.Then,a power line model growth and separation algorithm under KD tree organization is proposed,which can extract and separate a single power line point at one time.Then,using a reconstruction model combining straight line and parabola to reconstruct power line.Finally,a power grid safety distance calculation algorithm based on virtual grid is proposed to measure the safety distance of power line points and other ground points to realize the marking of dangerous points.The experiment proves that the detection process proposed in this paper obtains high power line model reconstruction accuracy and safety distance measurement accuracy,and the algorithm efficiency is high,which can meet the automatic inspection requirements of power to a certain extent.