Research On Plant Skeleton Extraction And Modeling Based On Point Cloud

Skeleton is a "compact" representation of the 3D model which abstractly reflect the topological structure and body characteristics of the plant model.The topology of the skeleton is the basic element of 3D reconstruction,retrieval and matching because it is simple and easy to operate.As a result,it is widely used in the fields of plant modeling,3D animation design,medical image and mapping image.The existing skeleton extraction algorithm,mostly using the human body and the object as the initial model,describes the 3D model based on the voxel data or the mesh surface information.However,there is few method of skeleton extraction which can directly work on the point cloud data of the plant.Due to the complex structure of plant morphology,the problem of self occlusion and the accuracy of the commonly used acquisition equipment,there is a lot of noise in the initial cloud data,as well as the missing of local data.If the traditional skeleton extraction method were adopted,it would be difficult to guarantee the correctness and completeness of the skeleton.Therefore,this paper presents a plant skeleton extraction method based on the Kinect equipment.The content of relevant research is as follows:(1)Using Kd-tree as the data structure to organize and manage the point cloud.Picking technology selects the required point cloud data from the three-dimensional plant model,as well as different colors serving as the initial mark to carry out the denoising and registration process of the scattered cloud point.(2)By searching the k-nearest neighbor of the target point to establish the topological relation of the point cloud.The k-means clustering method can be used to keep the branching extension function of the plant branch in the point cloud set to a great extent.Plant skeleton is built according to the angle relation of every target point and its adjacent point.(3)Analyzing and comparing the method of plant model reconstruction based on grid and point respectively.The real plant model is reconstructed by using the extracted plant skeleton as well as the surface structure of branches and leaves,which shows the shape and details of the original plant vividly.(4)By using the computer graphics theory of Open GL library programming,the prototype system is grouped with the function of interactive skeleton extraction,surface reconstruction,texture mapping in a modular way.In addition,our users set the relevant parameters as needed,which shows the experimental results more intuitively.