【作者】 张荣；

【导师】 伍永刚； 李衷怡；

【作者基本信息】 华中科技大学 ， 系统分析与集成， 2007， 硕士

【摘要】 随着激光测距扫描等三维空间数据获取硬件技术的不断完善,人们可以获得精度和密度都越来越高的物体表面三维数据,利用物体表面三维数据来建立真实的物体数字模型。它也促进了计算机图形学,虚拟现实和多媒体通信技术的不断发展,与此同时,海量数据的处理也成为三维重建这一研究领域的热点之一。对海量断层数据集进行三维重建,它的主要处理难点是被处理的数据量庞大,对计算机的存储空间提出了很高的要求,同时在重建物体后的交互显示的时候,其绘制速度比较慢。普通计算机无法提供其所需的各类资源,通常不得不使用大容量的图形工作站才能解决问题。为了有效重建这类复杂实体,本文提出一种任务分配算法,即分段重建。通过对任务的数据量来估算重建计算量,然后将数据量按计算量大致相同分配到各个负载,实现分布式三维重建中计算资源的合理分配。我们针对虚拟中国人的部分切片数据,完成了表面重建工作,结果表明分段重建算法能在有限内存空间里完成传统重建算法不能完成的大数据量的曲面重建。这对海量数据重建工作,如医学图象可视化等应用起到一定帮助作用。因此,文章中提出的任务分配确实能够比较有效和灵活地解决海量数据的重建工作。更多还原

【Abstract】 In the wake of improvement of hardware technique of 3D dataset capturing such as laser ranging-scanning technique and so on,people may obtain surperficial dataset of object with more and more precision and density to depend on,utilize the dataset on the ground to establish the real object’s numeral model.It aslo help the development of computer graphics,visual reality,multimedia communication.At the same time,the technology of processing large scale slice dataset have become one of the hotspots in the field of 3D reconstruction.The main difficulty of processing large scale slice dataset is the large quantity of data,which has the high stroage requirment to the computer. It aslo may decrease the speed of drawing graphics during the interactive displaying,after reconstructing the objects.Ordinary computer cannot offer various resources,therefore,we have to use graphic station with enough memory space.To reconstruct such complex object efficiently,in this thesis a task-allocation.Algorithm is proposed,to reconstruct the task that has been divided. The task-allocation algorithm mentioned in the paper,can estimate the computational complexity through task’s quantity,and then equally assigns the task’s quantity to each load based on computational complexity, to properly allocates the resoures in distributed 3D reconstruction.We finish the reconstruction by part slice dataset from Visible Human Datasets.The results show that large surfaces reconstruction is viable in a personal computer uning ours algorithm instead of traditon method.Therefore,the algorithm proposed in the paper really can efficiently and conveniencely resolve the reconstruction of large scale slice datasets.更多还原