摘要
为了提高高光谱遥感影像的分类精度,提出了一种基于稀疏非负最小二乘编码的高光谱数据分类方法。采用非负最小二乘方法,将待测样本表示为训练样本的线性组合,并将得到的系数作为待测样本的特征向量,通过最小误差方法对待测样本进行分类。提出的方法在AVIRIS Indian Pines和萨利纳斯山谷高光谱遥感数据集上进行分类实验,并和主成分分析(PCA)、支持向量机(SVM)和基于稀疏表示分类器(SRC)方法进行比较,在2个数据集上本文方法的总体识别精度分别达到85.31%和99.56%,Kappa系数分别为0.816 3和0.986 7。实验结果表明本文方法的总体识别精度和Kappa系数都优于另外3种方法,是一种较好的高光谱遥感数据分类方法。
In order to improve the classification accuracy and reduce computation complexity, a hyperspectral remote sensing data classification method based on sparse non-negative least-squares coding was proposed. By adopting non-negative least-squares,the test samples were expressed as a linear combination of training samples,and the obtained coefficients were used as its feature vector. As a result of the non-negative constraint,the feature vectors were sparse,which can not only improve the efficiency of the proposed algorithm,but also enhance the discrimination performance of algorithm. At last,the minimizing residual was used to classify the test samples. The experimental verifications of the proposed method were carried out on AVIRIS Indian Pines and Salinas Valley hyperspectral remote sensing data,the classification accuracies of the proposed method were 85. 31% and 99. 56%, and the Kappa coefficients were 0. 816 3 and 0. 986 7,respectively. The proposed method was compared with PCA,SVM and SRC in terms of classification accuracy and Kappa coefficients on two databases,experiment results showed that the proposed method was superior to PCA,SVM and SRC. The proposed approach was valuable for hyperspectral data classification with low computational cost and high classification accuracy,it was a better method of hyperspectral remote sensing data classification.
出处
《农业机械学报》
EI
CAS
CSCD
北大核心
2016年第7期332-337,共6页
Transactions of the Chinese Society for Agricultural Machinery
基金
甘肃省自然科学基金项目(145RJZA183)
关键词
稀疏非负最小二乘
高光谱遥感
数据分类
sparse non-negative least-squares
hyperspectral remote sensing
data classification
