Abstract
Recently, convolutional neural networks (CNNs) have been used to extract spectral and spatial features of hyperspectral images (HSIs) for hyperspectral image classification (HSIC) because of their excellent performance in extracting and analyzing complex data. However, due to the limited labeled samples and existing mixed pixels, it is difficult to extract features effectively, which further leads to the problem of overfitting of the model. On the other hand, to improve the extraction ability of the CNN, the depth of the model, and the complexity of the convolution kernel often need to be increased. In this article, a sandwich CNN based on spectral feature enhancement (SFE-SCNN) is proposed for HSIC. The proposed method, SFE-SCNN, introduces the spectral feature enhancement operation, which makes the data reflect more discriminative spectral feature details to suppress the interference of mixed pixels. Furthermore, according to the preprocessed data structure features, a lightweight sandwich convolution neural network is proposed. To fully extract the spectral features, the spectral feature re-extraction operation is used for the first time. Experimental results on three real hyperspectral datasets demonstrate that the proposed method achieves better classification performance than other state-of-the-art methods.
Highlights
HYPERSPECTRAL images (HSIs) possess hundreds of continuous spectral bands so that it contains rich spectral information while containing spatial information [1,2]
hyperspectral image classification (HSIC) is based on the rich detail features of the HSIs in the spectral domain and the spatial characteristics of ground objects to achieve the classification of each pixel in the images
Different from the traditional way of data augmentation methods applied in the daily images, due to the 3-D characteristics of HSI, the spectral information can be structurally exchanged with the spatial information dimensions to realize spectral feature enhancement (SFE)
Summary
HYPERSPECTRAL images (HSIs) possess hundreds of continuous spectral bands so that it contains rich spectral information while containing spatial information [1,2]. HSIC is based on the rich detail features of the HSIs in the spectral domain and the spatial characteristics of ground objects to achieve the classification of each pixel in the images. To further improve the classification performance, some classification methods based on joint extraction of spectral-spatial features have been proposed gradually [14,15,16,17,18]. Different from the traditional way of data augmentation methods (i.e., flip, rotation, and noise) applied in the daily images, due to the 3-D characteristics of HSI, the spectral information can be structurally exchanged with the spatial information dimensions to realize spectral feature enhancement (SFE). A novel SCNN is proposed to extract the spectral-spatial features from processed HSI cubes It uses a series of small convolution kernels of different sizes to achieve multi-scale.
Talk to us
Join us for a 30 min session where you can share your feedback and ask us any queries you have