Research On Key Technologies Of Large-scale Data Real-time Rendering

3D graphics rendering is widely used in many fields,such as game entertainment,virtual reality,scientific computing visualization,to meet real-time and real rendering.In recent years,the computing performance of software and hardware has been dramatically improved.Still,with the increase of large-data,real-time rendering is facing enormous challenges.This article mainly focuses on the real and real-time rendering technologies of volume data and geographic terrain data.For volume data with high accuracy and large volume,3D visualization is computationally intensive.Meanwhile,the volume illumination calculation accentuated the computational burden.And it has always been a problem how to process and render large-scale geographic terrain data efficiently.Simultaneously,anti-aliasing technologies can significantly enhance the visual effect in data rendering,reduce image aliasing and flicker,and bring additional rendering overhead.The data amount growth has undoubtedly increased the computational overhead of various rendering technologies,even though parallel rendering technologies that improve rendering performance still have many challenges in surface rendering and volume rendering.In response to the above technical problems,the main work and innovations of this article include the following:(1)Volume data parallel rendering and lighting calculationThe performance scalability research is carried out to solve a large amount of illumination calculation based on two mainstream rendering methods of slicing and ray casting.In order to solve the problem that the slice-based volume rendering of multiple rendering passes is inefficient in calculating global illumination,a multi-slice per pass(MSPP)algorithm is proposed to improve the computing performance from the data and algorithm parallel.With the same number of slices,the MSPP algorithm can improve the half-angle slicing algorithm's performance by about twice.Aiming at the problem that the complex sampling of ray casting volume rendering leads to a large amount of illumination calculation,a slice-based ray casting(SBRC)method is proposed to improve the rendering efficiency and effect.First,the lighting information of the entire geometry slices of the volume data is rendered slice by slice with the light source as the viewpoint to the light attenuation buffer.Secondly,in the ray casting process,the light attenuation buffer is used to calculate the sampling point's lighting influence,including the calculation of volume shadows,soft shadows,and scattering.The SBRC algorithm only needs the time overhead of one pass and realizes rendering performance scalability by varying the number of slices and each slice's resolution.Experiments show that the above two methods can significantly improve the rendering efficiency and rendering effect of volume data and meet scalable volume rendering requirements.(2)The high-efficiency organization and rendering of large-scale geographic terrain data3D geographic terrain rendering mainly includes two main processes: geographic data organization and 3D terrain construction.To maximize the performance of terrain rendering and support Level of Detail(LOD)and Mipmap technologies to improve performance,a flexible,seamless division strategy of digital elevation model with consistent latitude and longitude range of digital orthophotos is proposed.Each terrain tile is organized into components,segments,and quadrilateral grids to quickly generate real 3D terrain with efficient LOD scheduling.Experiments show that this method can promptly build scalable 3D terrain and reduce draw calls.(3)Anti-aliasing technologies in data renderingAiming at the problem that the hardware anti-aliasing methods are not directly compatible when they are turned on in the deferred shading stage,a sub-pixel reconstruction continuous edges(SRCE)method is proposed.First,multi-sampling anti-aliasing is turned on,and the scene is rendered to the geometry buffer(G-Buffer).Secondly,Chebyshev's inequality is used to detect edge pixels on sub-pixels through probability statistics and to identify edge pixels and ordinary pixels to reduce the consistency overhead of shading threads.Thirdly,the continuous edge is reconstructed from the sub-pixels by the #filter method.Finally,edge pixels are adaptively shaded for anti-aliasing processing.SRCE minimizes the "staircase" effect on the edges of 3D objects.Moreover,it can also be combined with post-processing or time antialiasing technologies to enhance the anti-aliasing effect.Aiming at the incompatibility between deferred shading technologies and coverage sample anti-aliasing(CSAA),a large scene anti-aliasing rendering framework based on deferred shading technology is proposed.The framework can use different anti-aliasing levels for different rendering objects to reduce delay and stutter and balance the effect and efficiency.Experiments show that these methods can effectively improve the quality of data rendering while balancing rendering performance.(4)Large-scale data parallel rendering virtualized frameworkBased on mainstream parallel rendering framework research,combined with virtualization technology,a virtualized framework for 3D graphics parallel rendering is designed to improve the computing power rendering capabilities of the large-scale data in volume rendering,surface rendering,and anti-aliasing technologies.The Equalizer is used as the parallel rendering middleware to organize more flexible parallel rendering function modules and virtualize various calculating resources as rendering units for the schedule and manage on-demand.In summary,this article studies a series of critical technical issues of real-time rendering of volume data and geographic terrain data,thoroughly improving the real and real-time rendering capabilities and effectively improving the practicability of large-scale data rendering through data parallelism,algorithm parallelism,and computing power parallelism.