Study And Implementation Of Real-time Visualization Techniques For Realistic Large-scale Terrain In Flight Simulator

Through simulating the feelings during flight, including visual, audio, motion, force etc, the flight simulator can build a realistic flight environment on the ground. So it has an important role in pilot training etc. With the development of our aviation industry and the launch of the big aircraft project, the study of key techniques for flight simulator is especially emergent. Three dimensional terrain is the important component of the virtual scene outside cockpit, and the realism and real-time of terrain visualization have the direct impact on the fidelity, immersion and effectiveness of the flight simulation. For the application requirements of flight simulator and characteristics of terrain model, the visualization system is required to interactively render a realistic large-scale outdoor terrain scene with a high rate. The contradiction between the massive datasets and current hardware has become the bottleneck for system implementation, and will be exacerbated with the increasing demands for model accuracy and rendering performance. So based on the features of flight simulation and hardware, the study of this paper focuses on real-time visualization technology of the realistic large-scale terrain, and will provide a real-time terrain visualization solution for the flight simulator.Realism and real-time are the two main objectives, and real-time is the premise. So with the limited hardware, the contradiction of visualization system is mainly reflected in hareware processing speed, data storage, scheduling and transmission capacity. So the real-time visualization technology in this paper involves massive data rendering acceleration, layout and schedule method, progressive compression. And for enhancing the realism, the real-time controllable detail synthesis is studied.Firstly, for massive datasets rendering acceleration, the regular-chunk based smooth adaptive terrain rendering algorithm is studied. For GPU data processing, the regular-chunk based model structure and the adaptive rendering framework integrated with visibility culling are given. And for the regular chunk structure, the refining criterion based on view-frustum triangle density and rendering optimization method based on template are adopted. Furthermore the integrated solution for space and time discontinuity problem is proposed to make sure the smooth terrain rendering. Results show that the algorithm can adapt GPU's characteristics, the rendering efficiency is improved efficiently; and can realize the smooth massive terrain datasets rendering.Secondly, to overcome the memory limitation, out-of-core technology based on chunk hierarchy is explored, the layout and schedule method for massive data are study focuses.And the objective is reducing the impact of data scheduling on system performance. On the basis of data relativity analysis relate to data acquisition mode, terrain cluster construction method and cluster-based chunk layout way are given to arrange the data chunks that will be continuously gotten together. Data scheduling use the priority-based scheduling mechanism and data prefetch is introdued. For prefetch efficiency, based on conservative view-frustum, the calculation method for predicted nodes and their priority is proposed. Results show that the page faults and data loading time is reduced effectively, and the scheduling efficiency is improved.Furthermore, to reduce the pressure of massive data storage and transmission, based on lifting wavelet, an algorithm integrating terrain data progressive compression and real-time adaptive mesh construction is proposed. The decompression efficiency and GPU-oriented implementation are research focuses. The lifting wavelet transform model for terrain chunk is built to map the chunk simplification and refinement to corresponding wavelet transforms. And then the chunk quadtree is constructed using the chunk-based wavelet transform. The wavelet coefficients generated are progressively compressed by an indirect look-up table technique. Based on the chunk-hiberarchy compressed, the active chunks for a viewport are decided, and the datasets needed are incrementally loaded and uncompressed meantime. The real-time chunk data decompression and wavelet transforms are achieved on GPU. Results show that the algorithm can compress datasets efficiently, the decompression speed is improved, and the real-time performance for visulization is enhanced.To increase the details and reality of terrain scene and reduce its dependence on massive high-resolution datasets, a method for chunk-based real-time controllable detail synthesis is proposed. For the regular chunk, four-point subdivision is exploited to generate new samples for extending the chunk and a perlin-noise based controllable multifractal algorithm is presented to compute the sample disturbance. For disturbance computation, the influencing factors for terrain fractal are analyzed and an impact factor is introduced to make sure the detail synthesis is controlled by real terrain features. The new chunks synthesized together with original chunks form a two-level adaptive chunk hierarchy, and terrain rendering is completed. Results show that the algorithm can real-time produce fine large terrain scene similar to the true terrain using limited terrain samples.Finally, terrain rendering system integrated with above key techniques for real-time visualization of large-scale terrain is built and applied to the PC-based flight simulator, and interactive large terrain fly-through driven by flight simulation or manual manipulation is realized. The practicality and effectiveness of the large-scale terrain visualization program are verified.