The fusion of hyperspectral and light detection and range (LiDAR) data has been an active research topic. Existing fusion methods have ignored the high-dimensionality and redundancy challenges in hyperspectral images (HSIs), despite that band selection methods have been intensively studied for HSI processing. This article addresses this significant gap by introducing a cross-attention mechanism from the transformer architecture for the selection of HSI bands guided by LiDAR data. LiDAR provides high-resolution vertical structural information, which can be useful in distinguishing different types of land cover that may have similar spectral signatures but different structural profiles. In our approach, the LiDAR data are used as the “query” to search and identify the “key” from the HSI to choose the most pertinent bands for LiDAR. This method ensures that the selected HSI bands drastically reduce redundancy and computational requirements while working optimally with the LiDAR data. Extensive experiments have been undertaken on three paired HSI and LiDAR datasets: Houston 2013, Trento, and MUUFL. The results highlight the superiority of the cross-attention mechanism, underlining the enhanced classification accuracy of the identified HSI bands when fused with the LiDAR features. The results also show that the use of fewer bands combined with LiDAR surpasses the performance of state-of-the-art fusion models.

Hyperspectral remote sensing sensors have the ability to capture a wide range of spectrum of ground objects with hundreds to thousands of bands. The obtained hyperspectral images contain more detailed spectral information than conventional panchromatic or color images. However, redundant and noisy data could also be introduced into the images and may impair further processing of the data. Removing the redundancy and noisy bands becomes one of the most important preprocessing steps of hyperspectral imaging. As a feature selection method, an attention mechanism has been successfully used in computer vision and natural language processing to enable algorithms to concentrate on the most significant data. In this article, we propose a band attention network for hyperspectral image classification. This network can automatically learn to attend to the desired band set that maximizes the classification accuracy. With careful design, the band attention framework can undertake both band weighting and band selection tasks. When working in the band weighting mode, the proposed band attention framework can learn a band weighting vector that models the relationship between all the bands. For band selection, our network can be adapted as two different types of band selection networks that select the significant bands and discard the useless bands. In both tasks, the attention and classification can be learned end to end. The experimental results prove the effectiveness and advantages of the proposed work.

Band selection is an effective means to alleviate the curse of dimensionality in hyperspectral data. Many methods select a compact and low redundant band subset, which is inadequate as it may degrade the classification performance. Instead, more emphasis shall be put on selecting representative bands. In this article, we propose a robust unsupervised band selection method to address this issue. Our method reveals bandwise representativeness based on the comprehensive interband neighborhood structure. It incorporates an interband neighborhood graph into a sparse self-contained regression model in order to provide a reasonable measure for bandwise representativeness. The derived coefficient matrix not only uncovers bandwise importance values but also is coherent to the generalized interband local neighborhood structure. For constructing the interband neighboring structural graph, an integrated multigraph model is employed to achieve better generalization performance. It combines the benefit of multiple graphs but is insusceptible to the defects of a single one. To enhance the reliability of this model, a joint trace minimum and nonnegative constraint is imposed on the coefficient matrix. Accordingly, a multigraph integrated embedding and robust self-contained regression model (MGRSR) is formulated. In addition, an iterative update algorithm is developed to solve the problem. Comparative experiments on three hyperspectral data sets illustrate that MGRSR is robust to various data and has superior performance compared with several state-of-the-art methods.

Deep learning-based methods have greatly improved the performance of hyperspectral image classification over the past several years. Nevertheless, current deep learning methods require training and deploying an independent model for each hyperspectral data domain. Representations learned for one data domain can hardly be generalized to other data domains, so multiple models are needed in real-world applications when data from multiple domains are involved. In this paper, we design a single neural network that learns universal representations simultaneously from multiple hyperspectral remote sensing data domains. The universal convolutional neural network adapts its behaviour to different hyperspectral datasets. The majority of parameters of the network are shared to learn common knowledge from multiple datasets. A small number of domain-specific parameters are assigned to handle the domain shift. In addition, we propose a two-step training strategy to fully utilize the capacity of the universal network. Experiments conducted on seven hyperspectral image datasets demonstrate that the proposed universal network outperforms multiple individual specialized single domain networks.