中国物理B ›› 2022, Vol. 31 ›› Issue (12): 120701-120701.doi: 10.1088/1674-1056/ac8cd7

• • 上一篇    下一篇

Learnable three-dimensional Gabor convolutional network with global affinity attention for hyperspectral image classification

Hai-Zhu Pan(潘海珠)1,†, Mo-Qi Liu(刘沫岐)1, Hai-Miao Ge(葛海淼)1, and Qi Yuan(袁琪)2   

  1. 1 College of Computer and Control Engineering, Qiqihar University, Qiqihar 161000, China;
    2 College of Telecommunication and Electronic Engineering, Qiqihar University, Qiqihar 161000, China
  • 收稿日期:2022-04-11 修回日期:2022-08-26 接受日期:2022-08-26 出版日期:2022-11-11 发布日期:2022-11-19
  • 通讯作者: Hai-Zhu Pan E-mail:panhaizhu@qqhru.edu.cn
  • 基金资助:
    Project supported by the Fundamental Research Funds in Heilongjiang Provincial Universities (Grant No. 145109218) and the Natural Science Foundation of Heilongjiang Province of China (Grant No. LH2020F050).

Learnable three-dimensional Gabor convolutional network with global affinity attention for hyperspectral image classification

Hai-Zhu Pan(潘海珠)1,†, Mo-Qi Liu(刘沫岐)1, Hai-Miao Ge(葛海淼)1, and Qi Yuan(袁琪)2   

  1. 1 College of Computer and Control Engineering, Qiqihar University, Qiqihar 161000, China;
    2 College of Telecommunication and Electronic Engineering, Qiqihar University, Qiqihar 161000, China
  • Received:2022-04-11 Revised:2022-08-26 Accepted:2022-08-26 Online:2022-11-11 Published:2022-11-19
  • Contact: Hai-Zhu Pan E-mail:panhaizhu@qqhru.edu.cn
  • Supported by:
    Project supported by the Fundamental Research Funds in Heilongjiang Provincial Universities (Grant No. 145109218) and the Natural Science Foundation of Heilongjiang Province of China (Grant No. LH2020F050).

RichHTML

0

PDF (PC)

75

赞

摘要： Benefiting from the development of hyperspectral imaging technology, hyperspectral image (HSI) classification has become a valuable direction in remote sensing image processing. Recently, researchers have found a connection between convolutional neural networks (CNNs) and Gabor filters. Therefore, some Gabor-based CNN methods have been proposed for HSI classification. However, most Gabor-based CNN methods still manually generate Gabor filters whose parameters are empirically set and remain unchanged during the CNN learning process. Moreover, these methods require patch cubes as network inputs. Such patch cubes may contain interference pixels, which will negatively affect the classification results. To address these problems, in this paper, we propose a learnable three-dimensional (3D) Gabor convolutional network with global affinity attention for HSI classification. More precisely, the learnable 3D Gabor convolution kernel is constructed by the 3D Gabor filter, which can be learned and updated during the training process. Furthermore, spatial and spectral global affinity attention modules are introduced to capture more discriminative features between spatial locations and spectral bands in the patch cube, thus alleviating the interfering pixels problem. Experimental results on three well-known HSI datasets (including two natural crop scenarios and one urban scenario) have demonstrated that the proposed network can achieve powerful classification performance and outperforms widely used machine-learning-based and deep-learning-based methods.

关键词: image processing, remote sensing, 3D Gabor filter, neural networks, global affinity attention

Abstract: Benefiting from the development of hyperspectral imaging technology, hyperspectral image (HSI) classification has become a valuable direction in remote sensing image processing. Recently, researchers have found a connection between convolutional neural networks (CNNs) and Gabor filters. Therefore, some Gabor-based CNN methods have been proposed for HSI classification. However, most Gabor-based CNN methods still manually generate Gabor filters whose parameters are empirically set and remain unchanged during the CNN learning process. Moreover, these methods require patch cubes as network inputs. Such patch cubes may contain interference pixels, which will negatively affect the classification results. To address these problems, in this paper, we propose a learnable three-dimensional (3D) Gabor convolutional network with global affinity attention for HSI classification. More precisely, the learnable 3D Gabor convolution kernel is constructed by the 3D Gabor filter, which can be learned and updated during the training process. Furthermore, spatial and spectral global affinity attention modules are introduced to capture more discriminative features between spatial locations and spectral bands in the patch cube, thus alleviating the interfering pixels problem. Experimental results on three well-known HSI datasets (including two natural crop scenarios and one urban scenario) have demonstrated that the proposed network can achieve powerful classification performance and outperforms widely used machine-learning-based and deep-learning-based methods.

Key words: image processing, remote sensing, 3D Gabor filter, neural networks, global affinity attention

中图分类号:  (Neural networks, fuzzy logic, artificial intelligence)

  • 07.05.Mh
93.85.Pq (Remote sensing in exploration geophysics) 93.90.+y (Other topics in geophysical observations, instrumentation, and techniques) 95.75.Mn (Image processing (including source extraction))