综合多特征的Landsat 8时序遥感图像棉花分类方法

发布时间：2018-02-28 05:41

本文关键词： 多时相Landsat 数据综合多特征 NDVI 棉花提取特征选择　出处：《遥感学报》2017年01期 　论文类型：期刊论文

【摘要】：传统的多时相遥感图像分类大多拘泥于单一特征,本文基于多时相的Landsat 8遥感数据,开展了综合多特征的特征提取与特征选择方法研究。综合了NDVI时间序列、最佳时相反射率光谱特征以及纹理特征作为初始分类特征,并采用基于属性重要度的粗糙集特征选择算法对其进行特征约简。分类结果表明:(1)利用初始分类特征,分类的总体精度达到92.81%,棉花提取精度达87.4%,与仅利用NDVI时间序列相比,精度分别提高5.53%和5.05%;(2)利用粗糙集选择后的特征分类,分类总体精度可达93.66%,棉花分类精度达92.73%,与初始分类特征提取结果相比,棉花分类精度提高5.33%。基于属性重要度的粗糙集特征选择不仅提高了分类精度,同时有效降低了分类器的计算复杂度。
[Abstract]:The traditional classification of multitemporal remote sensing images is mostly confined to a single feature. Based on the multi-temporal Landsat 8 remote sensing data, the feature extraction and feature selection methods of multi-temporal remote sensing images are studied. The NDVI time series are synthesized. The optimal temporal reflectance spectral feature and texture feature are used as the initial classification features, and the feature selection algorithm based on attribute importance is used to reduce the feature. The classification results show that: 1) the initial classification feature is used. The overall accuracy of classification is 92.81, and the precision of cotton extraction is 87.4. Compared with only using NDVI time series, the accuracy is increased by 5.53% and 5.05 respectively. The classification accuracy of cotton is 93.666.The precision of cotton classification is 92.733.Compared with the result of initial classification feature extraction, the classification accuracy of cotton is increased by 5.33.The rough set feature selection based on attribute importance not only improves the classification accuracy, but also improves the classification accuracy of cotton. At the same time, the computational complexity of the classifier is reduced effectively.
【作者单位】：中国科学院遥感与数字地球研究所;中国矿业大学环境与测绘学院;
【基金】：高分辨对地观测系统重大专项(编号:11-Y20A40-9002-15/17)~~
【分类号】：TP751

【参考文献】