Research On Extracting Skeleton Of Tree Point Cloud Based On Cylindrical Prior

Point cloud is a discrete point set in three-dimensional space,and the curve skeleton of point cloud is the abstract representation of point cloud.Curve skeletons extracted from tree point clouds play an important role in the reconstruction of biological and structural models of trees.For example,the tree trunk and branches can be reconstructed by taking the curve skeleton as the axis.However,when there are noise points and occlusion in the point cloud of the tree,the curve skeleton generated by traditional methods may not be consistent with the topological structure of the point cloud,or disconnected.Therefore,to overcome these problems,this thesis proposes a new method,which is used to extract topologically correct and well-centered tree skeletons from tree point clouds.The main work of this thesis is summarized as:①This thesis deeply studies the relevant theoretical knowledge of tree skeleton extraction,which provides a theoretical basis for the proposed method in this thesis.At the same time,some classical methods are introduced.②Downsampling the raw point clouds by octree.These problems exist in the raw point clouds:1)the point clouds may contain noise and outliers;2)there can be large amounts of missing data in point clouds due to data occlusions;3)the point density in the point clouds can vary substantially;4)The number of data in the point cloud is too large.In order to solve problems 1)and 2),this thesis uses an octree to store the raw point cloud,and downsamples the data points in each leaf node of octree to obtain the noise-free point cloud with uniform density.③Extracting the initial curve skeleton.After obtaining the point cloud by octreebased downsampling,this thesis first constructs the k-nearest-neighbor graph over the point cloud,then constructs the level set based on the k-nearest-neighbor graph,and finally extracts the initial curve skeleton from the level set.④Optimizing skeleton points.The extracted initial curve skeleton is poorly centered and the topology structure is not accurate enough.Based on the prior knowledge that the distances from each point of the cross section on the trunk to the axis of symmetry of the cylinder are similar,this thesis proposes a cylindrical prior constraint(CPC)method to optimize the centroid of the skeleton points.Then the radius of the branch is estimated with the help of Da Vinci’s formula,and the unsuitable bifurcation points are corrected.Finally,the curve skeleton with good centeredness and accurate topological structure is obtained.⑤Evaluating skeleton extraction algorithm.A standard tree point cloud data set is constructed.The data set includes four types:1)containing data missing;2)uneven density distribution;3)containing noisy points;4)different number of data points.Then a quantitative evaluation method is proposed:Skeleton Point Deviation(SPD),which is used to quantitatively evaluate the centeredness of skeleton points.By comparing quantitative evaluation and visual effects,the curve skeleton extracted by this method is better than the traditional method in terms of topological accuracy and centeredness.