基于GPU的海量城市管线高效建模与实时绘制

Efficient Modeling and Real-Time Rendering of Massive Urban Pipelines Based on GPU

地理信息系统行业积累了海量基于二维矢量的管线数据,文中针对传统的基于CPU的多层次细节预处理三维建模及绘制方法存在质量和效率方面的不足,从管线二维矢量数据的形状特征出发,提出一种无需预处理的、可一次性在GPU中装载并处理城市规模管线数据的三维建模与实时绘制算法.该算法利用现代GPU可编程硬件的特性来实现,在GPU上对管线形状进行解码,在顶点着色器采用2个剪枝策略进行场景的可见性剔除计算,然后基于细分着色器对管线曲面进行多层次细节几何自动建模,全过程无需CPU干预.实验结果表明,文中算法适用于城市级别的海量管网数据,相比于已有的方法,绘制效果和多项性能指标得到了大幅提升.

Large amount of 2D vector pipeline data accumulates in the GIS industry, however, the traditional CPU-based LOD preprocesses of 3D modeling and rendering have deficiency in both quality and efficiency, which requires to be improved urgently. Based on the structural feature of original data, we present a novel one-time initialization method for modeling and rendering, which requires no preprocessing steps or block-by-block loading. The whole algorithm thoroughly exploits the programmable characteristics of modern GPU. The shape of the pipeline is decoded with GPU. Visibility-culling of the scene is conducted on vertex shad-er using two pruning strategies. Then LODs automatic geometric modeling is performed on tessellation shader. The entire process is executed without making use of the CPU.Result shows that the algorithm we proposed can be applied to massive pipeline data at the city level. Compared with traditional LOD preprocessing methods, improvements can be found in both rendering outcomes and several performance indicators.