GPU-Based Rendering and Collision Simulation of Ground Vegetation in Large-Scale Virtual Scenarios
PubDate: December 2019
Teams: Universidade Federal de Santa Maria
Writers: Flavio Paulus Franzin; Cesar Tadeu Pozzer; Bruno Torres do Nascimento
The user’s immersion is highly related to the visual aspects of virtual scenarios. Thus, it is imperative for such applications the rendering of ground vegetation with high fidelity. However, the representation of these elements requires a high storage and graphics processing, which are the bottleneck of the other approaches. We present a GPU-based solution for handling dense ground vegetation on large-scale scenarios. Our proposal includes an architecture to distribute, render, and, to deal collisions with dynamic objects. The plants are placed in the scenario based on positions computed in real-time. Collisions are encoded in vector fields, which store the deformations to be applied to the plants during the rendering. These collisions may result in temporary or permanent deformation on the plants. The rendering performance is optimized through LOD and Instancing techniques, and the memory cost is reduced compressing the vector fields. The results show that the proposed approach is capable of handling ground vegetation for large-scale scenarios, ensuring a pleasant visual with low processing and storage costs.