摘要
提出了灰度共现矩阵类特征基的特征提取算法,该算法中不同类型的纹理具有与之相对应的灰度共现矩阵特征基,使得在某一类纹理的灰度共现矩阵特征基下,同类样本的特征值最大,该特征既反映了不同类纹理灰度共现矩阵的相似性又反映了不同类纹理灰度共现矩阵间距,本文以中国科学院电子学研究所机载合成孔径雷达图像进行实验,结果表明,该方法不仅使分类过程简单,而且分类精度明显提高.
A feature extracting algorithm of class characteristic base computed from Gray Level Cooccurrence matrix is proposed, in this method different texture corresponds to different Gray Level Cooccurrence matrix feature base which maxizes the feature value of similar spaciman, this feature reflects similarity and distance of Gray Level Co-Occurrence matrix from different texture. The test is executed with airborne Synthetic Aperture Radar （SAR） image from Institute of Electronics, Chinese Academy of Sciences （IECAS）, the results show this algorithm makes classifing process simple and classifing accuracy obviously improved.
出处
《测试技术学报》
EI
2005年第3期310-314,共5页
Journal of Test and Measurement Technology
关键词
灰度共现矩阵
纹理
合成孔径雷达
地物分类
Gray Level Go-Occurrence Matrix （GLCM）
texture
SAR
classification of things on ground
