Key Techniques For Real-time 3d Interactive Visualization Of Massive Data

With the explosive growth of the scientific data, a major challenge that should be faced is how to effectively understand and use these vast amounts of data. In order to solve this problem using visual analysis, some key techniques for real-time 3D interactive visualization of massive data, including 3D modeling, compression and coding, data transmission, interactive rendering, 3D animation display are investigated in this thesis.The huge amounts of data generated by photolithography simulation and the unorganized point cloud data obtained through 3D scanner are regarded as raw data for 3D modeling in this work. Firstly, triangular mesh models and point based models are reconstructed from this raw data. Then these 3D models are compressed and coded. With management and transferring of the massive data, the interactive rendering of 3D scene is realized. Finally, the deformation and movement of object in virtual scene are displayed by 3D animation. The main contributions of this work are as follows:Firstly, this thesis proposes a 3D modeling algorithm for photolithography simulation based imaging. 3D models are constructed from the data of light intensity distribution generated by photolithography imaging simulation.Secondly, a perceptual coding algorithm for 3D model based on wavelet subdivision surfaces is presented. This thesis focuses on the 3D models reconstructed from depth map and the models are compressed according to characteristics of human visual perception. In this way, high-quality 3D model of perception can be coded with less data.Thirdly, a joint mesh and texture coding method is introduced. This work combines the wavelet subdivision surface with the image wavelet transformation. Therefore, the texture image and the mesh with same level of detail can be synchronously transmitted and displayed. Fourthly, this thesis proposes massive data blocks prediction and dynamic loading algorithm. In order to solve the problem that the massive data can't be loaded into memory as a whole, this work divides the mass of data into smaller blocks. With real-time prediction and dynamic loading of data blocks, the large 3D scene can be displayed with interaction in real-time.Fifthly, a hybrid point and polygon rendering method based on data blocks is introduced. In order to achieve high rendering quality with progressive transmission, this work builds point models and polygon models according to different features of geometric blocks respectively. Then, the 3D scene is rendered by hybrid method by making use of both advantages of point and polygon.Finally, the deformation algorithm for photo-resist during the photolithography is introduced. With the lithography development model, the position of vertices at each time during development can be determined. Then, the mesh describing the photo-resist surface is constructed with these vertices. Thus, with merging and splitting the vertices during the simulation of development, high-quality 3D animation of the photo-resist surface deformation can be achieved.Experimental results show that the algorithms proposed in the thesis achieved the satisfactory performance and provided useful approaches to real-time 3D interactive visualization for massive scientific data.