摘要
提出了一种新的基于自相关的高光谱遥感图像纹理特征提取算法,该算法通过引入核函数技术,将单波段纹理窗口的空间自相关函数,扩展到多波段遥感图像的纹理描述。然后对特征矢量进行无监督C均值聚类实验和有监督RBF神经网络分类实验,在分类实验中确定了最佳窗口尺寸。实验结果表明,该文提出的自相关特征可以有效地描述高光谱遥感图像的纹理。
A novel texture feature extraction method of hyperspectral remote sensing image was proposed based on autocorrdation function. The concept of space autocorrelation function was expanded form texture window of single-band image to that of multi-band remote sensing image by introducing kernel function technique. The optimal size of texture window was determined during classification experiments. The results of C-means clustering and RBF neural networks classification experiments show that, the proposed texture feature based on autocorrelation can effectively describe the texture of hyperspectral remote sensing image.
出处
《武汉理工大学学报》
CAS
CSCD
北大核心
2009年第3期10-13,17,共5页
Journal of Wuhan University of Technology
基金
国家自然科学基金(40371079)
关键词
高光谱遥感
纹理
特征提取
hyperspectral remote sensing
texture
feature extraction
