小波包信息熵特征矢量光谱角高光谱影像分类

发布时间：2018-06-16 16:32

本文选题：信息熵 + 小波包子频段　；参考：《中国图象图形学报》2017年02期

【摘要】：目的针对高光谱数据波段多、数据存在冗余的特点,将小波包信息熵特征引入到高光谱遥感分类中。方法通过对光谱曲线进行小波包分解变换,定义了小波包信息熵特征矢量光谱角分类方法(WPE-SAM),基于USGS光谱库中4种矿物光谱数据的分析表明,WPE-SAM可增大类间地物的可区分性。在特征矢量空间对Salina高光谱影像进行分类计算,并讨论了小波包最佳分解层的确定,分析了WPE-SAM与光谱角制图(SAM)方法的分类精度。结果 Salina数据实例计算表明:小波包信息熵矢量能较好地描述原始光谱特征,WPE-SAM分类方法可行,总体分类精度(OA)由SAM的78.62%提高到WPE-SAM的78.66%,Kappa系数由0.769 0增加到0.769 5,平均分类精度(AA)由83.14%提高到84.18%。此外,通过Pavia数据验证了WPE-SAM分类方法具有较强的普适性。结论小波包信息熵特征可较好地表示原始光谱波峰、波谷等特征信息,定义的小波包信息熵特征矢量光谱角分类方法(WPE-SAM)可增大类间地物可区分性,有利于分类。实验结果表明,WPE-SAM分类方法技术可行,总体精度及Kappa系数较SAM有一定的提高,且有较强的普适性。但WPE-SAM方法精度与效率有待进一步提高。
[Abstract]:Aim to introduce the feature of wavelet packet information entropy into hyperspectral remote sensing classification. Methods based on the wavelet packet decomposition transformation of the spectral curve, the wavelet packet information entropy characteristic vector spectral angle classification method was defined. Based on the analysis of four mineral spectral data in the USGS spectral database, it was shown that WPE-SAM could increase the distinguishing ability of the ground objects among classes. Salina hyperspectral images are classified and calculated in feature vector space. The determination of optimal decomposition layer of wavelet packet is discussed. The classification accuracy of WPE-SAM and spectral angle mapping method is analyzed. Results the calculation of Salina data shows that the wavelet packet information entropy vector can well describe the original spectral features and the WPE-SAM classification method is feasible. The total classification accuracy increased from 78.62% of SAM to 78.66kappa coefficient of WPE-SAM from 0.769 to 0.769 5, and the average classification accuracy increased from 83.14% to 84.18%. In addition, the WPE-SAM classification method is verified by Pavia data. Conclusion the wavelet packet information entropy features can well represent the original spectral peaks and troughs. The defined wavelet packet information entropy feature vector spectral angle classification method (WPE-SAM) can increase the distinguishability of ground objects among classes and is beneficial to classification. The experimental results show that the WPE-SAM classification method is feasible, and the overall precision and Kappa coefficient are higher than that of SAM. However, the accuracy and efficiency of the WPE-SAM method need to be further improved.
【作者单位】：中国矿业大学(北京)地球科学与测绘工程学院;安徽理工大学测绘学院;
【基金】：国家自然科学基金项目(41271436)~~
【分类号】：TP751

【参考文献】