遥感图像波谱角并行分类算法  

Algorithm of Spectrial Angle Parallel Classification on Remote Sensing Image

作　　者：刘晓云[1] 康一梅[1] 齐同军[2] 方金云[2] 

机构地区：[1]北京航空航天大学软件学院,北京100191 [2]中科院计算技术研究所,北京100190

出　　处：《计算机科学》2009年第9期267-270,共4页Computer Science

摘　　要：针对海量遥感数据处理提出了一种遥感图像波谱角并行分类算法。该算法是基于单机波谱角算法。针对其处理遥感图像数据速度慢、效果差、内存不足等缺点,提出了在集群环境下遥感图像波谱角并行分类算法的同步、互斥和负载均衡等策略,以及在多台设备上并行处理的方法。通过对算法的时间复杂度、加速比进行分析,并在集群环境下进行实例验证,将分类图像与知名软件ENVI进行对比,验证了算法的优越性和有效性。This paper proposed a spectral angle parallel classification algorithm on remote sensing image against mass remote sensing data processing. Based on single-machine spectral angle algorithm, aiming at defects in processing remote sensing image data, such as slow speed, poor effect, and insufficient internal memory, this algorithm proposed some synchronized, exclusive and load balancing strategies against spectral angle parallel classification algorithm on remote sensing image, as well as method of parallel processing on multiple equipments under cluster environment. The paper compared classified image with famous software ENVI through analyzing time complexity and speedup ratio of the algorithm and doing verification by actual examples under the cluster environment to inspect and verify the superiority and validity of the algorithm.

关 键 词：并行处理 波谱角分类 遥感图像分类 

分 类 号：TP751.1[自动化与计算机技术—检测技术与自动化装置]