一种基于网格计算的农业遥感数据快速分类算法  

A agricultural remote sensing data rapid classification algorithm based on grid computing

作　　者：张超[1] 李建成[1] 王剑秦[2] 

机构地区：[1]西安工程大学计算机科学学院,陕西西安710048 [2]中国农业大学信息与电气工程学院,北京100083

出　　处：《西安工程大学学报》2010年第6期810-813,817,共5页Journal of Xi’an Polytechnic University

基　　金：陕西省教育厅专项科研计划项目(07JK251)

摘　　要：针对农业遥感数据分类速度慢的问题,本文对K-means算法的并行化方法进行了研究,提出了一种基于网格计算的快速分类算法.在实现K-means并行算法过程中,通过调整通信数据的存储方式,使该算法每次迭代的通信次数降低为一次全收集操作.该算法节点数的选取是根据数据规模和分类需求确定的,而不是节点数越多算法效率越高.实验结果表明,这种方法具有较好的加速比.It is very slow in the classification of remote sensing data in agriculture.After analyzing K-means algorithm and its parallel methods,a rapid classification algorithm based on grid computing is proposed.The algorithm adjusting the data storage structure during achieving communication for parallel K-means algorithm,the communication type between nodes changed to all-gather methods,by which,in each iteration,the parallel K-means algorithm only requires one communication.The experimental results show that the method has a good speedup,and also show that it is not more efficient as the higher number of nodes,the number of selected nodes should be determined by the data scale and classification requirements.

关 键 词：遥感 K-MEANS算法 并行算法 网格计算 

分 类 号：TP391[自动化与计算机技术—计算机应用技术]