融合空间信息的高光谱影像稀疏表达分类

Sparse Representation Classification of Hyperspectral Image Based on Spatial Information Fusion

针对高光谱影像空谱信息利用问题,设计了一种融合空间信息的稀疏表达分类方法,以提高高光谱影像的分类精度。首先,在特征提取阶段引入空间信息,采用形态学滤波的方法提取高光谱影像的形态学属性剖面特征;然后,采用训练样本构成的字典对提取到的空间特征进行稀疏编码,在编码过程中进一步引入空间邻域信息来提高稀疏编码效果;最后,根据测试样本的稀疏编码向量计算其相对于每个类别的重构误差,并将该样本划分到重构误差最小的类别中完成分类。为了验证该方法的有效性,在Pavia大学和Indian pines 2组高光谱数据集上进行分类实验。实验结果表明,该方法充分利用了高光谱影像的空间邻域信息,能够有效提高高光谱影像的分类精度。

Aiming at the problem of spatial spectrum information utilization of hyperspectral images,a sparse representation classification method integrating spatial information is designed to improve the classification accuracy of hyperspectral images.Firstly,spatial information is introduced in the feature extraction stage,and the morphological attribute profile features of hyperspectral images are extracted by morphological filtering.Then,the dictionary composed of training samples is used to sparse encode the extracted spatial features,and the spatial neighborhood information is further introduced into the coding process to improve the sparse coding effect.Finally,the reconstruction error relative to each category is calculated according to the sparse coding vector of the test sample,and the sample is divided into the category with the smallest reconstruction error to complete the classification.In order to verify the effectiveness of this method,classification experiments are carried out on two sets of hyperspectral data sets of Pavia university and Indian pines.The experimental results show that this method makes full use of the spatial neighborhood information of hyperspectral images and can effectively improve the classification accuracy of hyperspectral images.