基于小波变换和多尺度注意力机制的高光谱图像分类模型
Hyperspectral Image Classification Model Based on Wavelet Transform and Multi-Scale Attention
DOI: 10.12677/jisp.2025.142015, PDF,    科研立项经费支持
作者: 易守恒：西安石油大学电子工程学院，陕西 西安；焦 龙*：西安石油大学化学化工学院，陕西 西安
关键词: 小波变换；多尺度注意力机制；高光谱图像分类；Wavelet Transform； Multi-Scale Attention； Hyperspectral Image Classification
摘要: 高光谱图像(Hyperspectral Image, HSI)分类旨在利用高光谱数据丰富的光谱信息，实现细粒度物质识别。传统的卷积神经网络(Convolutional Neural Network, CNN)在提取空间特征方面表现出了卓越的能力，但往往难以有效捕捉光谱特征中固有的顺序和层次关系，从而限制了其在HSI分类任务中的性能。为了解决这一挑战，我们提出了一种新的模型，该模型将小波变换(Wavelet Transform, WT)与多尺度注意机制(Multi-Scale Attention, MSA)相结合，以提升分类性能。具体来说，WT能够有效地将光谱空间信息分解为多个频率子带，便于细粒度特征的提取和分层分析；MSA会结合不同大小的卷积核来提取图像的不同尺度信息，从而更好地捕捉局部特征和全局特征的关系。通过学习不同尺度的重要特征，实现多尺度信息的高效融合。在公开的高光谱数据集上的实验结果表明，该模型在分类精度方面显著优于传统的基于CNN模型和Transformer模型。这些结果充分表明，将WT与MSA相结合，能够在高光谱图像分类的任务中展现出巨大的潜力。
Abstract: Hyperspectral Image (HSI) classification aims to leverage the rich spectral information in hyperspectral data for fine-grained material identification. Traditional Convolutional Neural Networks (CNNs) have shown exceptional ability in extracting spatial features but often struggle to effectively capture the inherent order and hierarchical relationships in spectral features, thus limiting their performance in HSI classification tasks. To address this challenge, we propose a novel model that combines Wavelet Transform (WT) with Multi-Scale Attention (MSA) to enhance classification performance. Specifically, WT can effectively decompose spectral-spatial information into multiple frequency subbands, facilitating the extraction of fine-grained features and hierarchical analysis. MSA, on the other hand, combines convolution kernels of different sizes to extract multi-scale information from the image, thereby better capturing the relationships between local and global features. By learning the importance of features at different scales, MSA enables efficient fusion of multi-scale information. Experimental results on publicly available hyperspectral datasets show that this model significantly outperforms traditional CNN-based models and Transformer models in classification accuracy. These results fully demonstrate that the combination of WT and MSA holds great potential in hyperspectral image classification tasks.
文章引用：易守恒, 焦龙. 基于小波变换和多尺度注意力机制的高光谱图像分类模型[J]. 图像与信号处理, 2025, 14(2): 149-161. https://doi.org/10.12677/jisp.2025.142015

参考文献

[1] Freitas, S., Silva, H., & Silva, E. (2022). Hyperspectral imaging zero-shot learning for remote marine litter detection and classification. Remote Sensing, 14(21), 5516.
[2] Yuan, J., Wang, S., Wu, C., & Xu, Y. (2022). Fine-grained classification of urban functional zones and landscape pattern analysis using hyperspectral satellite imagery: A case study of Wuhan. IEEE Journal of Selected Topics in Applied Earth Observations and Remote Sensing, 15, 3972-3991.
[3] Wang, C., Liu, B., Liu, L., Zhu, Y., Hou, J., Liu, P., & Li, X. (2021). A review of deep learning used in the hyperspectral image analysis for agriculture. Artificial Intelligence Review, 54(7), 5205-5253.
[4] Kumar, D., & Jha, R. (2024, March). Identifying Rocks and Mineral Resources Using Hyper Spectral Analysis. In 2024 2nd International Conference on Artificial Intelligence and Machine Learning Applications Theme: Healthcare and Internet of Things (AIMLA) (pp. 1-6). IEEE.
[5] Ma, L., Crawford, M.M., & Tian, J. (2010). Local Manifold Learning-Based $k$-Nearest-Neighbor for Hyperspectral Image Classification. IEEE Transactions on Geoscience and Remote Sensing, 48, 4099-4109.
[6] Fauvel, M., Benediktsson, J. A., Chanussot, J., & Sveinsson, J. R. (2008). Spectral and spatial classification of hyperspectral data using SVMs and morphological profiles. IEEE Transactions on Geoscience and Remote Sensing, 46(11), 3804-3814.
[7] Ham, J., Chen, Y., Crawford, M. M., & Ghosh, J. (2005). Investigation of the random forest framework for classification of hyperspectral data. IEEE Transactions on Geoscience and Remote Sensing, 43(3), 492-501.
[8] Huang, X., & Zhang, L. (2012). An SVM ensemble approach combining spectral, structural, and semantic features for the classification of high-resolution remotely sensed imagery. IEEE transactions on geoscience and remote sensing, 51(1), 257-272.
[9] Fang, L., Yan, Y., Yue, J., & Deng, Y. (2023). Towards the vectorization of hyperspectral imagery. IEEE Transactions on Geoscience and Remote Sensing.
[10] Ge, Z., Cao, G., Li, X., & Fu, P. (2020). Hyperspectral image classification method based on 2D-3D CNN and multibranch feature fusion. IEEE Journal of Selected Topics in Applied Earth Observations and Remote Sensing, 13, 5776-5788.
[11] Zhou, W., Kamata, S. I., Luo, Z., & Wang, H. (2021). Multiscanning strategy-based recurrent neural network for hyperspectral image classification. IEEE Transactions on Geoscience and Remote Sensing, 60, 1-18.
[12] Liang, H., Bao, W., Shen, X., & Zhang, X. (2021). Spectral-spatial attention feature extraction for hyperspectral image classification based on generative adversarial network. IEEE Journal of Selected Topics in Applied Earth Observations and Remote Sensing, 14, 10017-10032.
[13] Chen, Y., Jiang, H., Li, C., Jia, X., & Ghamisi, P. (2016). Deep feature extraction and classification of hyperspectral images based on convolutional neural networks. IEEE transactions on geoscience and remote sensing, 54(10), 6232-6251.
[14] Haut, J. M., Paoletti, M. E., Plaza, J., Plaza, A., & Li, J. (2019). Visual attention-driven hyperspectral image classification. IEEE transactions on geoscience and remote sensing, 57(10), 8065-8080.
[15] Zhou, F., Hang, R., Liu, Q., & Yuan, X. (2019). Hyperspectral image classification using spectral-spatial LSTMs. Neurocomputing, 328, 39-47.
[16] Vaswani, A. (2017). Attention is all you need. Advances in Neural Information Processing Systems.
[17] Alexey, D. (2020). An image is worth 16x16 words: Transformers for image recognition at scale. arxiv preprint arxiv: 2010.11929.
[18] Hong, D., Han, Z., Yao, J., Gao, L., Zhang, B., Plaza, A., & Chanussot, J. (2021). SpectralFormer: Rethinking hyperspectral image classification with transformers. IEEE Transactions on Geoscience and Remote Sensing, 60, 1-15.
[19] He, X., Chen, Y., & Lin, Z. (2021). Spatial-spectral transformer for hyperspectral image classification. Remote Sensing, 13(3), 498.
[20] Mou, L., & Zhu, X. X. (2019). Learning to pay attention on spectral domain: A spectral attention module-based convolutional network for hyperspectral image classification. IEEE Transactions on Geoscience and Remote Sensing, 58(1), 110-122.