谐波能量谱特征向量的高光谱影像Bayes分类

发布时间：2018-04-11 08:19

本文选题：高光谱影像 + 频率域变换　；参考：《计算机应用研究》2017年05期

【摘要】：对于高光谱影像存在高维非线性、数据冗余多、纯训练样本难以提取等不足,引入频率域空间的谐波分析(harmonic analysis,HA)理论并提出了一种高光谱影像的HA-Bayes监督分类方法。该方法在保持高光谱数据空—谱特性不变的情况下,从光谱维角度分析不同分解层的影像光谱谐波特征,将高光谱影像变换成由谐波能量谱组成的频率域特征矢量信息。通过建立谐波能量谱特征向量的先验知识,实现Bayes准则下谐波能量谱特征矢量信息判别与分类,最终实现高光谱影像分类。将此方法应用到ROSIS高光谱影像分类时获得的分类总体精度达85.5%,Kappa系数也达到了0.812。进一步实验也证明了频率域的谐波分析在高光谱遥感影像特征提取与分类方面具有更好的优势和潜力。
[Abstract]:For hyperspectral images with high dimensional nonlinearity, high data redundancy and difficulty in extracting pure training samples, the theory of harmonic analysis in frequency domain is introduced and a HA-Bayes supervised classification method for hyperspectral images is proposed.Under the condition of keeping the space-spectrum characteristics of hyperspectral data unchanged, the spectral harmonic characteristics of different decomposition layers are analyzed from the perspective of spectral dimension, and the hyperspectral image is transformed into the characteristic vector information in frequency domain composed of harmonic energy spectrum.By establishing the prior knowledge of the characteristic vector of harmonic energy spectrum, we can distinguish and classify the characteristic vector information of harmonic energy spectrum under Bayes criterion, and finally realize the classification of hyperspectral image.When this method is applied to the classification of ROSIS hyperspectral images, the overall accuracy of the classification is up to 85. 5% and the coefficient of Kappa is 0. 812.Further experiments also prove that harmonic analysis in frequency domain has better advantages and potential in feature extraction and classification of hyperspectral remote sensing images.
【作者单位】：中国矿业大学(北京)地球科学与测绘工程学院;
【基金】：国家自然科学基金资助项目(41271436) 中央高校基本科研业务费专项资金资助项目(2009QD02)
【分类号】：TP751

【相似文献】