基于改进的面向对象遥感影像分类方法研究——以西藏米林县典型林区为例

发布时间：2018-02-09 02:50

本文关键词： 西藏面向对象最优分割尺度最近邻法阈值分类法　出处：《遥感技术与应用》2017年03期 　论文类型：期刊论文

【摘要】：针对目前面向对象多尺度分类方法中,最优分割尺度的确定方法不具有普适性或者易受主观性影响的问题,以西藏米林县的Landsat 8OLI影像为数据源,对研究区影像多尺度分类进行研究。首先确定多尺度分类的最优分割尺度,提出基于多尺度分类精度Kappa系数的最优分割尺度函数模型法,在此基础上,利用多尺度分类分别与最近邻分类和阈值分类法相结合的方法,对研究区影像进行分类。结果表明:分割尺度分别为190、150、100、60,多尺度分类法比单一尺度分类精度高;最近邻多尺度分类法比阈值多尺度分类精度高,其总精度分别为86%和72%,Kappa系数分别为0.72和0.69。最优分割尺度函数模型在具有普适性的基础上更具有科学理论性,多尺度分类与最邻近分类结合的方法比与阈值分类结合的方法分类效果好,为后续植被动态变化监测提供了依据。
[Abstract]:In view of the problem that the method of determining the optimal segmentation scale is not universal or subject to subjectivity in the current object-oriented multi-scale classification methods, the Landsat 8OLI image in Milin County, Tibet, is used as the data source. Firstly, the optimal segmentation scale of multi-scale classification is determined, and the optimal segmentation scale function model based on Kappa coefficient of multi-scale classification accuracy is proposed. The method of combining multi-scale classification with nearest neighbor classification and threshold classification method is used to classify the image of the study area. The results show that the segmentation scale is 190 ~ 150 ~ 100 ~ (60), and the accuracy of multi-scale classification is higher than that of a single scale classification. The accuracy of the nearest neighbor multiscale classification is higher than that of the threshold multiscale classification. The total accuracy of the nearest neighbor multiscale classification is 86% and 72kappa coefficient is 0.72 and 0.69 respectively. The optimal partition scale function model is more scientific and theoretical on the basis of universality. The combination of multi-scale classification and nearest neighbor classification is more effective than that of threshold classification, which provides a basis for monitoring vegetation dynamic change.
【作者单位】：中国林业科学研究院资源信息研究所;西藏自治区林业调查规划研究院;
【基金】：国家科技基础性工作专项(2013FY111600)“中国森林植被调查”
【分类号】：S771.8;TP751

【相似文献】