Real-time physically-based animation of highly detailed grass

Abstract

Throughout the years, the rendering of a realistic nature scene has proven to be quite a challenge. The difficulties not only lay in the geometric and visual complexity of such scenes, but also in the description of the animation of each individual object. The large scale at which Mother Nature’s members are present, do not allow for a realistic animated nature scene to be acquired easily. The main contribution of this thesis is the presentation of an algorithm suitable for the physically-based-based animation of hundred of thousands blades of grass. The proposed animation algorithm deforms each blade individually while still preserving the ability to depict the scene in realtime. In addition, nearly all necessary calculations are computed by the GPU, which minimises the CPU’s workload. Furthermore, each blade is presented by a highly detailed geometric description. This, however, greatly complicates the animation process. The presented solution utilises self-generated keyframes in combination with data clustering and a number of interpolating techniques. The result is a rendering mechanism which allows the animation of each individual blade of grass. Considering all aspect of the animation algorithm, the author is convinced of the usability of the proposed techniques in the field of natural phenomena. The acquired results are most promising, and considering the necessary computation, perhaps even impressive.