Tree-Shaped 3D Model Three-Dimensional Reconstruction Based On The Topology Structure And Volume Rendering Method

The reconstruction of tree-shaped 3D models is mainly used in two fields.One is in the field of virtual reality to construct tree-shaped 3D models with complex and diverse topological structures.This has always been pursued by industries such as movies and games,and the other is in the field of medical imaging.For the reconstruction of some medical tree-like three-dimensional organ models,their general requirement is more accurate imaging results.The focus of this article is to improve the two mainstream methods in these two fields,namely the topology-based 3D reconstruction method commonly used in virtual reality and the volume rendering method commonly used in medical imaging.The improved method in this paper makes the tree-shaped 3D model generated by the topology-based 3D model reconstruction method better in the generation of small branches such as end branches,and makes the segmentation accuracy of the target region in the volume rendering 3D reconstruction method higher,And can intelligently generate the target medical tree-shaped three-dimensional organ model.In the 3D reconstruction process of the tree-shaped 3D model based on the topology structure,we use the interpolation in the tree model space to obtain a series of tree models,so that the generated tree-shaped model is compared with the traditional one to reduce the difficulty of complex modeling.The cumulative model combination method has more randomly generated tree models,which reduces the time to generate models compared to traditional manual 3D reconstruction.We propose the ELRand Perm Cauchy algorithm based on the tree model Rand Perm Cauchy algorithm in the tree model space,which can make the spanning tree model better in processing the details of the end branches and other small branches,closer to the same kind of tree model.In the 3D reconstruction process based on volume rendering,we use neural network to segment the input pixels,and then use the ray projection algorithm to perform 3D reconstruction of the slices after the pixel segmentation.This is an intelligent modeling process.Compared with traditional direct volume rendering,experts with prior knowledge of medicine and computer are required to participate.Joining neural network for pixel segmentation allows doctors to devote more energy to analyzing pathology..In the process of segmentation using neural networks,we proposed a new Multi Res SwinUnet network model to improve the accuracy of pixel segmentation for the tree-shaped organ model.In the step of the ray casting algorithm,we proposed a tree-like 3D model preprocessing algorithm based on the growth of the eight-neighborhood seed region.Since the results of the neural network segmentation are not accurate,a preprocessing process needs to be added before the ray casting algorithm.Compared with the traditional seed region growth algorithm,the accuracy of the rendering result is that the preprocessing algorithm of the tree-like 3D model based on the growth of the eightneighborhood seed region does not require manual interaction and can realize intelligent model reconstruction.Our two algorithm improvements improve the quality of the reconstructed model in the topology-based 3D reconstruction,and realize the intelligent model reconstruction process in the volume rendering-based 3D reconstruction,and improve the reconstruction quality of the target model.