基于贝叶斯分类的海量点集多核D-TIN并行算法  

The D-TIN parallel algorithm with multi-core of massive point set based on Bayesian classification

作　　者：乔梁[1] 

机构地区：[1]广东松山职业技术学院,广东韶关512126

出　　处：《信息技术》2017年第1期34-38,共5页Information Technology

基　　金：广东省科学技术发展基金(2009KC090)

摘　　要：针对当前海量点集数据构建D-TIN时所面临的数据量巨大且计算复杂耗时长的问题,研究提出一种基于朴素贝叶斯分类的并行算法。该算法以Delaunay三角网的空圆法则为基础,朴素贝叶斯分类后重组为理论依据,通过生成一组由确定三角形组成的三角形条带将点云划分为若干子域,然后对各子集进行并行D-TIN生成。通过对6GB的点云的D-TIN生成进行数值仿真,结果表明:该算法耗时6min,峰值内存占用仅为500MB,加速比为3.2,且执行过程中各处理器独立运行,无需互相通讯和同步。Due to the large amounts of data and the complex computations which consume much time,the current method of constructing D-TIN by very large point set has problems. This paper proposes a parallel algorithm which is based on Bayesian classification. The theoretical basis of this algorithm is the Delaunay principle and naive Bayes classifier. Firstly,the point cloud is divided into some subdomains by the triangle strips which consists of determined triangles. Then the subdomains can be used for parallel D-TIN generation. The result of this algorithm test the actual data shows that it only costed less time to generate D-TIN by 6GB point cloud and the peaking of memory footprint is 500 MB,also the speed up ratio is 3. 2. There is no communication or synchronization between processors while this algorithm is processing.

关 键 词：贝叶斯分类 DELAUNAY三角网 云点 并行计算 

分 类 号：TP391.1[自动化与计算机技术—计算机应用技术]