The aim of hyperspectral image (HSI) classification is to define categories for the different labels that are assigned to each pixel vector. Generative adversarial network (GAN) can mitigate the limited training sample dilemma to some extent, but there are still two critical issues, namely balance collapse and insufficient sample diversity. In this article, we propose a Coupled Dual-Channel Generative Adversarial Network for HSI classification. It mainly consists of a Coupled Generative Network (CGN) and a Dual-Channel Discriminative Network (DDN). CGN achieves the reconstruction of HSI samples through cascaded convolutional layers, in which the label information of the samples is used to avoid the balance collapse, while DDN extracts spatial attention weights and spectral attention weights of the input true/false samples respectively, and performs feature mining in both spatial and spectral dimensions in a dual-channel style. In order to reinforce the detailed features of input samples in both spatial and spectral dimensions, we design a new Cascaded Spatial-Spectral Attention Block (CSSAB). Finally, feature maps at different scales are fused for final sample discrimination and classification, which can mitigate the effects of insufficient sample diversity. Experimental results on two HSI data sets demonstrate that the proposed CDGAN effectively improves the classification performance compared to some state-ofthe-art GAN-based methods
High-resolution remote sensing images (HRRSIs) contain rich local spatial information and long-distance location dependence, which play an important role in semantic segmentation tasks and have received more and more research attention. However, HRRSIs often exhibit large intraclass variance and small interclass variance due to the diversity and complexity of ground objects, thereby bringing great challenges to a semantic segmentation task. In most networks, there are numerous small-scale object omissions and large-scale object fragmentations in the segmentation results because of insufficient local feature extraction and low global information utilization. A network cascaded by convolution neural network and global–local attention transformer is proposed called CNN-transformer cascade network. First, convolution blocks and global–local attention transformer blocks are used to extract multiscale local features and long-range location information, respectively. Then a multilevel channel attention integration block is designed to fuse geometric features and semantic features of different depths and revise the channel weights through the channel attention module to resist the interference of redundant information. Finally, the smoothness of the segmentation is improved through the implementation of upsampling using a deconvolution operation. We compare our method with several state-of-the-art methods on the ISPRS Vaihingen and Potsdam datasets. Experimental results show that our method can improve the integrity and independence of multiscale objects segmentation results.
Access to the requested content is limited to institutions that have purchased or subscribe to SPIE eBooks.
You are receiving this notice because your organization may not have SPIE eBooks access.*
*Shibboleth/Open Athens users─please
sign in
to access your institution's subscriptions.
To obtain this item, you may purchase the complete book in print or electronic format on
SPIE.org.
INSTITUTIONAL Select your institution to access the SPIE Digital Library.
PERSONAL Sign in with your SPIE account to access your personal subscriptions or to use specific features such as save to my library, sign up for alerts, save searches, etc.