Research On Some Key Technologies Of Modeling And Rendering Of Complex Scenes In Virtual Battlefield

Virtual battlefield has many advantages in military maneuver, such as scientific, economical, intuitional, and so on. Hence, it becomes an effective way that military countries carry on strategy, tactics simulation campaign drill. Reality is the main target that virtual battlefield is seeking for. With the expansion of the modern war and the unceasing enhancement of comprehensive, however, the battlefield environment has become more and more complex. Consequently, there are higher requirements on the simulation scale, the simulation precision, and the performance of interaction.To meet the demands of virtual battlefield applications, this dissertation focuses on the construction of virtual natural environment and the expression of artificial electromagnetic environment. Much effort is devoted to some key technologies of modeling and rendering of complex scenes in virtual battlefield, such as real-time rendering of large-scale terrain, modeling and dynamic simulation of grassland, modeling and rendering of cloud scene, and the three-dimensional expression of complex radar electromagnetic environment. Main contributions of this dissertation include:(1) To cope with the problems of visualization of very large-scale terrain in virtual battlefield, such as the requirement of large external memory and the difficulty of rendering large-scale terrain in real-time, a real-time rendering method for out-of-core terrain is proposed. In preprocessing stage, the full resolution terrain datasets are partitioned into a set of uniform blocks. For each block, a discrete hierarchy, which contains different resolution of height maps, is constructed. In order to save the space of hard disk, a new non-redundant data storage method in block is presented. The increase of rendering speed is achieved by using the view frustum culling and sight-line-dependent screen error computation method. To satisfy the continuity of navigation of terrain, a prefetching scheme based on view frustum extension is proposed, which implements the real-time loading of datasets. The experimental results show that our method achieves good visual effect with high FPS and provides complete solution for real-time rendering terrain scene with large dataset. (2) In view of the deficiency of traditional static expression of the vegetation scenes in battlefield environment, a method for dynamic simulation of grassland in wind is presented, which not only can realistically render the animated grass scenes under wind, but also can express the vibration effect of grass. Furthermore,a method for simulation of treading on grass is proposed. Use GPU-based real-time collision detection and stress vector transmission mechanism to compute deformation of grass in close view. The deformation computation of grass in moderate and far view is achieved by adjusting the gradient of billboard. It can not only effectively solve the problems of traditional methods, such as low simulation precision and the restricted shape of moving object, but also realistically simulate the interactive effect between battle unit and grass.(3) According to the problem of time consuming on clouds rendering, a method for the simulation of clouds is proposed. With this method, the Coupled Map Lattice (CML) is adopted for the modeling of cloud, and the simulation of scattering in clouds is achieved by using a series of spherical harmonics functions and coefficients to express incident-light distribution. A frequency domain volume rendering algorithm combined with spherical harmonics is used to implement fast rendering of cloud scenes.(4) To cope with the deficiency of traditional two-dimensional expression of the radar coverage in battlefield environment, a hybrid sampling method for the modeling of radar coverage is proposed. Based on the radar equation, the radar azimuth sampling step is determined by the resolution of terrain, so that the model has self-adaptability to different terrain resolution. The sub-area sampling of pitching angle based on the 3dB feature points, which ensures the model shape consistent with properties of radar while effectively reduces the number of drawing points. Furthermore, considering the environment factors on the radar detection range, a model-revised algorithm under terrain occlusion is presented based on geometrical optics. The experimental results show that the method can not only quickly complete computation, but also visually show the true radar detection range affected by complex terrain environment.