Abstract
Deep learning methods used for hyperspectral image (HSI) classification often achieve greater accuracy than traditional algorithms but require large numbers of training epochs. To simplify model structures and reduce their training epochs, an end-to-end deep learning framework incorporating a spectral-spatial cascaded 3D convolutional neural network (CNN) with a convolutional long short-term memory (CLSTM) network, called SSCC, is proposed herein for HSI classification. The SSCC framework employs cascaded 3D CNN to learn the spectral-spatial features of HSIs and uses the CLSTM network to extract sequence features. Residual connections are used in SSCC to accelerate model convergence, with the outputs of previous convolutional layers concatenated as inputs for subsequent layers. Moreover, the data augmentation, parametric rectified linear unit, dynamic learning rate, batch normalization, and regularization (including dropout and L2) methods are used to increase classification accuracy and prevent overfitting. These attributes allow the SSCC framework to achieve good performance for HSI classification within 20 epochs. Three well-known datasets including Indiana Pines, University of Pavia, and Pavia Center were employed to evaluate the classification performance of the proposed algorithm. The GF-5 dataset of Anxin County, obtained from China’s recently launched spaceborne Advanced Hyperspectral Imager, was also used for classification experiments. The experimental results demonstrate that the proposed SSCC framework achieves state-of-the-art performance with better training efficiency than other deep learning methods.
Highlights
Due to the recent development of hyperspectral remote sensing imaging technology, a large number of hyperspectral remote-sensing images with different spatial and spectral resolutions are available
We aimed to design an end-to-end spectral-spatial cascaded 3D convolutional neural network (CNN) with a convolutional long short-term memory (LSTM) (SSCC) framework motivated by the spectral–spatial residual network (SSRN), fast dense spectral–spatial convolution (FDSSC), and CLSTM
Cascaded 3D CNN extraction of spectral-spatial features from hyperspectral image (HSI) and CLSTM extraction of discriminative features from sequential image data are described in detail
Summary
Due to the recent development of hyperspectral remote sensing imaging technology, a large number of hyperspectral remote-sensing images with different spatial and spectral resolutions are available. Gong et al [25] fine-tuned pretraining models such as AlexNet and GoogleLeNet to capture the deep spatial features of HSI datasets and achieved high classification accuracy These methods focus on the extraction of HSI spatial information and fail to fully utilize the deep spectral-spatial features of HSIs. Corresponding to a 3D cube of the HSI, a three-dimensional convolutional neural network (3D CNN) is usually used to extract spectral–spatial features of HSIs effectively without any pre- or post-processing [9,11]. These models have low efficiency and slow convergence speeds, require numerous training epochs, and learn spatial information from HSIs in a manner that may introduce noise [24] To resolve these problems, we aimed to design an end-to-end spectral-spatial cascaded 3D CNN with a convolutional LSTM (SSCC) framework motivated by the SSRN, FDSSC, and CLSTM.
Talk to us
Join us for a 30 min session where you can share your feedback and ask us any queries you have