摘要
讨论了一种在集群计算环境下将物体各个方向的表面片拼接起来形成物体完整表面的算法通过采样计算出体素到每个表面片的最近距离,利用添加策略计算出每个体素的权重距离,再利用行进块算法抽取出物体表面实验结果表明,该算法在普通的PC集群上并行计算,可以大大减少建模的时间,特别是在大数据量时。
A volumetric view-merging algorithm to generate a consensus surface of an object from its range images under PC cluster is presented. After resampling the range surface, the signed distance for updating the voxel is computed. Then, the signed distance is combined with the data of voxel already acquired using a simple additive scheme. The surface mesh is extracted using the marching cube algorithm. Computation time for 1,2,4,8 PCs in parallel are recorded at two levels of shape approximation precision. The acceleration rate is nearly linear.
出处
《计算机辅助设计与图形学学报》
EI
CSCD
北大核心
2005年第5期895-901,共7页
Journal of Computer-Aided Design & Computer Graphics
基金
国家自然科学基金 (K60 0 75 0 0 5 )
关键词
3D表面重建
表面拼接
并行计算
集群计算
3D shape reconstruction
surface merging
parallel computing
cluster computing