Street Trees And Attribute Information Extraction From Vehicle-Borne Laser Point Cloud

Street trees are an important part of urban ecology and road systems.Traditional street tree information acquisition methods such as manual measurement,RTK measurement,and total station measurement are time-consuming and laborious,and the collected street tree information is single,which cannot meet the needs of the regulatory and planning departments to quickly collect and monitor street tree information.As an advanced measurement method,vehicle-borne mobile mapping system can quickly collect point cloud information of roads and their surrounding objects.The point cloud data obtained by the vehicle-borne mobile mapping system extracts the street tree and attribute information,and has the characteristics of low cost,high precision,strong timeliness,and all-weather operation.The spatial distribution of point cloud in the street tree is diverse and complex.Extracting the street tree and attribute information from the vehicle-borne laser point cloud data is a challenging subject.In this paper,the following research work is carried out around the problem of street tree and attribute information extraction in vehicle-borne laser point cloud.1.A ground filtering algorithm based on local region growth is proposed.Firstly,the algorithm establishes a regular grid for the point cloud space.Secondly,it uses the region segmentation to separate the ground points with obvious features and selects the reliable seed points in the ground point.Then,it uses the region plane fitting iterative calculation to optimize the seed point set.Finally,all ground points are extracted by local area growth to achieve accurate separation of ground points and non-ground points.The experimental part uses the algorithm to perform ground filtering and precision analysis on the two sets of urban road point cloud data,and the filtering effect is good,which verifies the effectiveness of the proposed method.2.A random forest algorithm is introduced to extract the street tree from the point cloud.The algorithm accurately separates the street tree from the point cloud data through the steps of multidimensional feature vector construction,feature selection,random forest model establishment and random forest classification.In the experimental part,the two sets of urban road point cloud data after ground filtering are used as data sources to test and analyze the accuracy.The results show that the algorithm can effectively extract street trees from the point cloud.The advantage of this algorithm is that through the steps of multi-dimensional feature vector construction and feature selection,it can fully select the features that play an active role in the classification of street trees.The application of multi-dimensional features can adapt to the data in complex environments,and the applicability is strong.3.Improved DBSCAN clustering algorithm.By analyzing the shortcomings of DBSCAN clustering algorithm,the algorithm is improved,and the improved clustering effect is improved,and the clustering time is reduced.The improved algorithm is applied to data denoising and point cloud clustering,and good results are achieved.4.Improved street tree segmentation and attribute extraction algorithms.Firstly,the improved DBSCAN clustering algorithm is used to cluster the street tree point cloud,and then the improved distance-based segmentation method is used to separate the connected trees.Finally,the attribute information of the street tree is obtained under the condition that part of the street tree point cloud is incomplete.The experimental part compares the attribute information extracted by human-computer interaction with the algorithm extraction result,and verifies the effectiveness of the method.