散度核协同表示与空谱融合特征的高光谱图像分类算法

A HYPERSPECTRAL IMAGE CLASSIFICATION ALGORITHM BASED ON DIVERGENCE KERNEL COLLABORATIVE REPRESENTATION AND SPATIAL-SPECTRAL FUSION FEATURE

协同表示的相关算法在高光谱图像分类的学习过程中没有很好地刻画高光谱图像的特性,如像素的局域性和标签信息等,因此其性能受到限制。对此,提出一种散度核协同表示技术并利用空谱融合获取特征的分类方法(IKCRC)。为有效刻画像素的局域性和标签信息,该方法构造新的散度核协同表示模型和相应的计算模型。在公式中引入核映射以提高分类能力,在计算过程上使用空谱融合的初步特征提取使得算法简洁高效。在标准高光谱图像数据集上进行的对比实验表明,IKCRC更能有效地提高分类精度。

In the learning process of hyperspectral image classification,the related algorithms of collaborative representation do not describe the characteristics of hyperspectral image well,such as pixel localization and label information,so its performance is limited.Therefore,we propose a divergence kernel collaborative representation technique and a classification method(IKCRC)that utilizes spatial-spectral fusion to obtain feature.In order to describe the localization and label information of pixels effectively,a new divergence kernel collaborative representation model and a corresponding computational model were constructed.In the proposed algorithm formula,we introduced kernel mapping to improve the ability of classification,and used the initial feature extraction of spatial-spectral fusion to make the algorithm simple and efficient.Comparison experiments on standard hyperspectral image data sets show that the proposed method(IKCRC)can improve classification accuracy more effectively.