Research On Real-time Rendering Technology Of Large-scale Infrared Scenes

Infrared scene rendering involves the calculation of temperature field,radiation field and interaction.The drawbacks of complex architecture,enormous computation have restricted its implementation in real-time rendering with the existing hardware.Therefore,aiming at the complex computation and time-consuming problem during the solution of background's temperature field,targets' radiation field and their interaction,this paper proposes a series of algorithms for fast computation and rendering,which remarkably improve the frame rate of real-time rendering,then introduces a GPU-based acceleration method.The specific work is as follows:A calculation model of background's temperature field based on neural network is established.The model for water surface,bare land surface,vegetation surface and man-made surface has been created respectively,utilizing the BP neural network optimized by genetic algorithm to calculate the temperature at any time of the day,with the factors which have great influence on surface temperature as inputs,such as solar radiation,temperature and wind speed etc.The results show that this method seems to calculate the temperature field of terrain backgrounds accurately.In order to render target more accurately and more quickly,this paper provides a target's radiation visualization algorithm based on texture mapping,which mapping the information of finite element model to texture of the geometric model.Therefore the drawing of dense finite element mesh is replaced by a geometric model drawing.This algorithm significantly reduces the number of patches in real-time rendering and improves rendering efficiency.On the other hand,a method based on pre-computation to calculate the reflection of targets is designed.To solve the time-consuming problem of BRDF calculation,this method stores the parameters calculation results in texture.A thermal shadow realization algorithm based on improved percentage closer soft shadow algorithm is propsed.By calculation and analysis on the temperature variation in thermal shadows,the filiter method of percentage closer soft shadow is improved.The algorithm not only deals with the problem that traditional method for visible band can't express temperature's gradient in thermal shadow,but also accelerates the realization of thermal shadow by GPU.Based on the research contents above,this paper designs a real-time rendering system for large-scale infrared scene,the efficiency of the algorithm is tested.Finally,the results show that this paper's method can render large-scale IR scene accurately and smoothly.