光子学报, 2014, 43 (8): 0810002, 网络出版: 2014-09-01
基于纹理特征和形态学特征融合的高光谱影像分类法
Based on Texture Feature and Extend Morphological Profile Fusion for Hyperspectral Image Classification
主成分分析 形态学特征 纹理特征 高光谱 灰度共生矩阵 Principle component analysis Extend morphological profile Texture feature Hyperspectral Gray level co-occurrence matrix
摘要
针对传统融合空间和光谱特征方法仅使用单一空间特征, 并未充分利用其双高分辨率的特点, 提出了一种基于纹理特征和形态学特征融合的高光谱影像分类方法.首先利用传统主成份分析变换降低高光谱影像的维数, 消除空间相关性, 然后对每一主成分采用灰度共生矩阵提取纹理特征, 获得扩展纹理特征, 最后结合形态学特征和部分光谱特征进行高光谱影像的分类.实验证明, 本文提出的方法能更好地克服传统光谱特征分类的局限性, 提高高光谱影像的分类准确度.
Abstract
Single spatial feature is used in the traditional spectral and spatial feature fusion,which does not make full use of the advantage of high spectral and spatial resolution.In order to overcome the shortage,a method based on texture feature and extend morphological profile fusion for hyperspectral image classification was proposed.Firstly,with the principle component analysis,the hyperspectral image dimension was reduced and the spatial correlation was eliminated,then using the gray level co-occurrence matrix the texture features for each principle component were extracted and the extend texture features were got,lastly combined the extend morphological profile and part spectral features hyperspectral image is classified.The experiments show that the proposed method can overcome the limitation of traditional spectral feature classification and improve the accuracy of hyperspectral images classification.
王增茂, 杜博, 张良培, 张乐飞. 基于纹理特征和形态学特征融合的高光谱影像分类法[J]. 光子学报, 2014, 43(8): 0810002. WANG Zeng-mao, DU Bo, ZHANG Liang-pei, ZHANG Le-fei. Based on Texture Feature and Extend Morphological Profile Fusion for Hyperspectral Image Classification[J]. ACTA PHOTONICA SINICA, 2014, 43(8): 0810002.