Research On The Planes Extraction Based On Laser Point Clouds

3D laser scanning technology with its unique advantages is widely used in engineering field.At present,the focus of 3D laser scanning technology is the point clouds data processing,according to the actual needs,extracting the useful information from a large number of disorder points has important value.In this paper,extracts the plane feature from the discrete point cloud as the starting point,study the normal vector estimation method,the plane feature extraction in the discrete point clouds data,the plane fitting method based on the zero threshold of plane curvature.The main works of the paper are as follows:1.When using local surface to estimate normal,although with good precision and similarity in the flat area,there is bad normal estimation in the transition or sharp feature surface,which is due to the different neighbor attributes.From this point,this paper improves a method of normal vector correction based on neighbor attribute information.First,using the eigenvalue method to get the initial normal vector,Secondly,given the different weight value by the spatial distance and normal vector deviation of each point in the neighbor,iterated to achieve the normal vector of the modified reference point.The experimental results show that the proposed method is effective and improves the accuracy of normal vector estimation.2.Analyzing the questions when using the regional growth and Hough Transform to extract plans,On this basis,a dynamic iteration cluster analysis is introduced.First,selecting the point with the smallest principal component,which closed to zero,as the initial class.Secondly,merge the neighbor points by the angle between points normal and distance between points.Then,calculate the zoom of the new class,iteration of the loop,Finally,merge the class that have less points.This method solve the problem that the plane of the same plane point cloud is divided into multiple planes when the plane is extracted,and can extract the plane feature point cloud in the discrete point clouds quickly and effectively.3.To solve the noise interference during the plan fitting,this paper proposes using the characteristic that points curvature on the plan close to zero value,extract the points with smaller curvature to fit the plan.This paper gives the way to set the zero value.Finally,the simulation results of the method are verified by simulation and real data.The experimental results show that the proposed method improves the accuracy of plane fitting in the same environment.