摘要
针对遥感影像数据量大的特点,设计并实现了一种针对海量遥感影像分类图的连通域标记算法。通过对影像进行合理高效的分块,解决海量数据处理过程中对大内存的要求。对读入内存中的数据,采用了基于队列的种子填充算法进行连通域标记,通过采用贯序连通域标记算法中的冲突表机制,解决因分块之间隔离而造成的同一连通域标记不同的问题。该方法具有可以处理海量数据、对分类结果数据仅仅访问一次、无须对生成的结果进行重新标记就可以完成连通域的标记、可以获取连通域的基本统计信息等特点。实验结果证明了该方法的高效性。
This paper designs and implements a connected component labelling method for remote sensing classification image, which is famous for its massive volume. The problem of huge EMS memory requirement for remote sensing data processing is resolved by an efficient partition scheme. The paper adopts a seed firing algorithm based on queues for connected component labelling in the memory; in order to clean up the label equivalence on the edges of connected blocks, a conflicted table is formed as in the sequential scan labelling algorithm. The method has the following advantage, it can process massive data, requires only one pass over the classification image, and it does not require any re-labelling mechanism to obtain basic statistic information for all connected component region. Experimentas results show that the method is efficient and effective.
出处
《计算机工程》
CAS
CSCD
北大核心
2008年第1期262-264,共3页
Computer Engineering
基金
国家"863"计划基金资助项目(2003AA135010)
关键词
海量遥感影像
连通域标记
分类后处理
massive remote sensing image
connected component labelling
post classification processing
