19 February 2025 in Expert Systems with Applications
14 December 2023 in IEEE Journal of Selected Topics in Applied Earth Observations and Remote Sensing
Due to their spatial and spectral information, hyperspectral images are frequently used in various scientific and industrial fields. Recent developments in hyperspectral image classification have revolved around the use of convolutional neural networks (CNNs) and transformers, which are capable of modeling local and global data. However, most of the backbone networks of existing methods are based on 3-D convolution, which have high complexity in network structure. Moreover, local information and global information are extracted through different modules, and the coupling relationship between the two types of information is weak. To address the above issues, we propose a method named main-sub transformer network with spectral-spatial separable convolution method (MST-SSSNet), which includes two key modules: the spectral-spatial separable convolution (SSSC) module and the main-sub transformer encoder (MST) module. The SSSC module uses the proposed spectral-spatial separable convolution, reducing network parameters and efficiently extracting local features. The MST module adds the designed sub-transformer in front of the conventional transformer encoder (main-transformer). It assists the main-transformer encoder to establish global correlation by learning local information. The WHU-Hi dataset can be used as a benchmark dataset for precise crop classification and hyperspectral image classification research. MST-SSSNet is shown to deliver better classification performance than current state-of-the-art methods on the datasets. The code will be downloaded to https://github.com/fengqinshou/MST-SSSNet .
15 January 2025 in International Journal of Applied Earth Observation and Geoinformation
27 August 2024 in IEEE Communications Letters
In an open electromagnetic environment, few-shot learning (FSL) has been widely used to recognize new classes of radar signals with few labeled data, which is needed in spectrum management or electronic reconnaissance systems for saving time and human resources. However, suffering from the distribution shift, existing methods minimizing the empirical risk on few labeled data may lead to model degradation on unseen data. We propose a Bayesian prototype learning (BPL) method for few-shot radar signal modulation recognition. Specifically, we design a Bayesian prototype (BP) learner that enhances generalization to unseen data, regularizing the prototype embedding space by modeling prototype learning as a variational inference problem. Furthermore, to transfer base class knowledge efficiently, we design a shallow-deep feature map fusion (SDF) block, combining feature maps from shallow and deep layers. Additionally, a class-covariance metric (CCM) is introduced to refine classification boundaries by considering intra-class distributions. Extensive experiments show the superiority of our method, achieving a recognition accuracy of 97.91% with 5 labeled data per class.
31 May 2025 in Signal Processing