Research And Application Of Level-of-Detail Technology Of Mesh Models

In Virtual Reality (VR), with the increase of scene complexity, the requirement of rendering image quality is getting higher and higher, which results in the rendering speed getting slower and slower. Image quality comes into conflict with rendering speed. However, real-time rendering is a basic requirement of VR system. The method of LOD (levels of detail) is an efficient measure to solve the confliction between scene complexity and rendering speed. LOD is a description method to reduce the number of polygons for rendering, storage, transmission and following processing by using different-resolution details for the objects in a scene, while ensuring the image quality or geometrical accuracy. Virtual Manufacturing (VM) is a branch of VR, it also requires solving the confliction between the scene complexity and processing speed. However, the models in VM are usually CAD models, so they possess their own characteristics. Curent LOD algorithms usually can not be used directly in VM, so the generation and rendering technologies of LOD are explored. New algorithms are proposed, and applied to CAD/CAM.Firstly, two 3-axis NC milling simulation methods are proposed using adaptive triangles. These two methods are derived from Lindstrom algorithm. Lindstrom algorithm is a LOD generation algorithm of regular mesh, it can generate the required LOD models efficiently and dynamically. However, the execution efficiency of Lindstrom algorithm can be further increased, an improved algorithm is given in this dissertation. On the basic of the improved algorithm, two NC milling simulation methods using adaptive discrete triangles are proposed, which are implemented with quadtree and binary tree respectively. These two methods can achieve the adaptive control of the density and number of triangles. Adaptive triangle methods can automatically adjust the density and size of triangles according to the fluctuation of the machining surface. The total number of triangles is greatly reduced, which greatly increases the efficiency of machining simulation. Not only is the image quality almost not affected, but also the workpiece model can be moved, rotated and zoomed during the process of simulation. When the inspections of machining error, overcut, undercut and interference are performed, the full resolution model in memory is used to ensure the accuracy and reliability of measurements and inspections.Secondly, the key features of mechanical CAD models are very important semantic elements. However, most of algorithms result in the loss of key features of models after simplification. In order to overcome this defect, a feature-preserved LOD algorithm is proposed. The process of mesh simplification is guided by the roughness of vertex and the deformation error of triangle after half-edge collapse. In addition, the estimation of vertex curvature of triangular mesh is proposed. A LOD generation algorithm is built based on the curvature and edge collapse using the feature of curvatures reflecting key features very well. Finally, models are required to transmit through network in many occasions, and the information security must be ensured. An effective transmission framework is proposed, which depends on progressive mesh. By using level-of-detail structure tree, the model with the most appropriate resolution is transmitted automatically according to the bandwidth and the requirement of users. Using this method, the data volume of transmission is reduced to the ultmost, and the resolution can be adjusted freely. With the combination of Christopher's role-based viewing envelopes for information protection, the information security in concurrent design environment is ensured, and the leakage risk of important data is reduced in the muti-organization cooperation environment.