中国虚拟人三维模型重建和可视化

3-D model reconstruction and visualization of virtual Chinese human

中国虚拟人项目已公开发布了数例人体数据集.与美国及韩国虚拟人数据集相比,中国虚拟人数据集提供的解剖和生理信息更为详尽.为解决数据量大处理复杂的难题,研究和实现了基于PC机环境的超大数据集三维模型重建和可视化.为提高重建精度,提出了几种处理算法:各向异性扩散预处理、相关性配准处理和基于水平集的分割处理.利用多处理器和多线程技术,提出并研究了任务并行、管线并行和数据并行的新并行策略,以及一种修正的同步模板算法,可有效提高处理速度.

The on going Virtual Chinese Human Project (VCHP) released some Chinese human datasets recently. Compared with US VHP datasets and Korea VHP datasets, the VCHP datasets provide more detailed information on anatomy and physiology. It is of critical importance to efficiently deal with the huge volume datasets in 3D model reconstruction and implementation of visualization on a PC-based system. For accurate reconstruction, some dedicated approaches are proposed, which are an anisotropic diffusion preprocessing, a correlation registration processing and a level set based segmentation processing. Taking advantage of multi-processor and multi-thread technique, a new parallel processing approach including task, pipeline and data parallelism and a modified synchronized templates algorithm are investigated and proposed, and the processing velocity can be effectively accelerated by this approach.