Resrarch On Tree Extraction Method In Outdoor Point Cloud Scene

In recent years,with the development of computer vision,various artificial intelligence technologies and robotics,the research on object classification of urban scenes has become a hot topic that researchers pay close attention to,and has been widely applied in the fields of robot perception navigation,personal safety and environmental monitoring,and automatic vehicle navigation.As one of the main elements of urban scenes,trees have important research significance in urban planning and extraction.However,due to the density of leaves,the diversity of tree species and the lack of data caused by occlusion,has no doubt increased the difficulty of trees element extracted from urban scene,therefore,research on tree extraction and segmentation in urban scenes is still a challenging frontier topic.To this end,this paper takes the single-sided point cloud data of the measured outdoor scene as the research object,and carries out the following research work:(1)An individual tree extraction method based on rough classification-fine segmentation is proposed.The method is based on the maximum feature set of the selected urban scene data,and divides the outdoor scene into "tree points" and "non-tree points" to obtain candidate trees in the scene,and then uses the fine segmentation strategy to extract an individual tree from the candidate trees,finally put forward a kind of weighted constraint rules to optimize a individual tree cluster,achieve the extraction of individual tree.(2)An individual tree extraction method based on shape classification and combination is proposed.The method firstly classifies the scene,that is,divides the scene into four categories:"scattered point","linear point","column point" and "planar point",and then on the basis of shape information for candidate canopy and candidate trunk,after use of the topological relationship between the crown and the trunk,the thought of integrated combination of shape,and combined with the proposed based on locating,filtering,matching and the method of optimization mechanism,realize the city extraction of single tree in the scene,so as to solve the crown block of the problems of overlap in the scene.(3)An individual tree extraction method based on energy function optimization is implemented.The method first obtains candidate trees by two classifications,and forms multiple tree clusters by clustering the candidate trees,then calculates the similarity between the tree clusters,and then uses the energy function minimization method to the trees with high similarity.The clusters are merged to realize the extraction of single tree clusters,and at the same time solve the problem of tree extraction errors caused by the presence of trees-like objects.