中文摘要：

多边形模型的布尔运算中包含复杂的求交计算以及多边形重建过程,精度控制和处理效率是其中的关键.为了降低布尔运算复杂度,提出一种适合硬件加速的基于渐进式布尔运算的多层次细节网格模型生成方法.该方法采用分层深度图像来近似表示多边形实体的封闭边界,将多边形的求交计算简化为坐标轴平行的采样点的实体内外部判断;为了免去各层次细节模型的重复采样过程,渐进式地将边界采样点归并到低分辨率下的立方体中;运用特征保持的多边形重建算法将相同立方体内的边界采样点转换成多边形顶点,根据邻接关系生成网格模型.上述算法使用支持图形硬件加速的CUDA编程并行实现.实验结果表明了算法的可行性.

英文摘要：

Boolean operations on polygonal models involve the complex intersection calculations and po- lygonal reconstruction, where the precision control and processing efficiency are two key problems. To re- duce the Boolean operation complexity, this paper proposes a progressive and GPU accelerated Boolean op- eration approach to generate levels-of-detail polygonal models. Layered depth images are employed to ap- proximate the enclosed boundaries of polygons and the intersection calculations are performed as the in/out classification of axis-aligned sampling points. To avoid the additional sampling process for levels-of-detail models, the boundary points are progressively merged into low-resolution cubes. The feature-preserving dual contouring algorithm is adopted to convert boundary points into a mesh model. The proposed algorithm can be implementation in parallel on GPU with the hardware-supported CUDA. Finally, experimental results show the feasibility of the proposed approach.