非等权距离改进的FCM多波段遥感影像监督分类算法

发布时间：2017-12-31 09:35

本文关键词：非等权距离改进的FCM多波段遥感影像监督分类算法　出处：《测绘地理信息》2016年04期 　论文类型：期刊论文

【摘要】：现实地理世界是一个极其庞大而复杂的多元系统,其间的众多的自然和半自然现象很难明确地划分为某种类别,在遥感影像中表现为混合像元的问题。因此,更符合遥感影像中地物特点的模糊聚类方法成为目前遥感分类领域的一个热点,用于有效解决混合像元分类的问题。但是,传统的模糊C均值聚类(FCM)算法仍然存在一定的局限性,例如,缺乏先验信息以及随机确定初始聚类中心等导致了分类结果精度不高。针对这些问题,提出一种改进的FCM算法,通过提取训练区并以训练区的光谱特性为基础改进FCM算法,以确定聚类中心,并实现各波段的非等权输入。最后以具有典型土地覆被状态的武汉市中心城区局部TM影像为例进行分类,研究表明,改进的FCM方法的分类精度显著高于经典的FCM算法,同时与传统硬划分方法迭代自组织数据分析技术ISODAT相比亦有显著优势;在效率方面,由于改进的FCM算法未采用自迭代方式,因此与经典的FCM算法相比,亦有明显提高。
[Abstract]:The real geographical world is an extremely large and complex multivariate system, in which a large number of natural and semi-natural phenomena are difficult to be clearly classified into a certain category, which is the problem of mixed pixels in remote sensing images. The fuzzy clustering method, which is more suitable to the features of ground objects in remote sensing images, has become a hot spot in the field of remote sensing classification, which is used to effectively solve the problem of mixed pixel classification. The traditional fuzzy C-means clustering algorithm still has some limitations, for example, the lack of prior information and random determination of the initial clustering center lead to the low accuracy of classification results. An improved FCM algorithm is proposed to determine the clustering center by extracting the training area and improving the FCM algorithm based on the spectral characteristics of the training area. Finally, the local TM image with typical land cover state is used as an example to classify the local TM image. The classification accuracy of the improved FCM method is significantly higher than that of the classical FCM algorithm, and it also has significant advantages compared with the traditional hard partitioning method, the iterative self-organizing data analysis technique ISODAT. In terms of efficiency, the improved FCM algorithm is not self-iterative, so compared with the classical FCM algorithm, it is also significantly improved.
【作者单位】：武汉大学资源与环境科学学院;中国人民大学公共管理学院;华中农业大学资源与环境学院;中南财经政法大学公共管理学院;香港中文大学地理与资源管理学系;
【基金】：国家基础科学人才培养基金“武汉大学地理科学理科基地”科研能力训练资助项目(J1103409)~~
【分类号】：P237
【正文快照】：

【参考文献】