全变差稀疏约束深度非负矩阵分解高光谱遥感影像解混方法

doi:10.16180/j.cnki.issn1007-7820.2023.02.008

摘要/Abstract

摘要：

传统非负矩阵分解方法仅基于单层线性模型,现有的深度非负矩阵分解模型忽略了地物光谱的实际混合物理过程,仅从数学理论考虑深度分解。对此,文中从光谱混合的物理过程出发,综合非负矩阵分解和深度学习,将光谱混合过程进行反向建模,并充分考虑丰度的稀疏性和空间平滑性,构建了用于高光谱遥感影像解混的面向端元矩阵的全变差稀疏约束深度非负矩阵分解模型。通过模拟实验和真实实验,将文中所提方法与5种解混方法进行对比。结果表明,相较于面向丰度的深度非负矩阵分解算法,文中所提方法的平均光谱角距离和均方根误差均有所降低,取得了最佳解混结果。

关键词: 高光谱遥感, 高光谱影像解混, 线性光谱解混, 非负矩阵分解, 深度学习, 深度非负矩阵分解, 稀疏约束, 全变差约束

Abstract:

The traditional nonnegative matrix factorization methods are based on single-layer model. The deep nonnegative matrix factorization methods are based on the mathematical theory, ignoring the actual spectral mixing process of materials. In this regard, this study starts from the physical process of spectral mixing, integrates non-negative matrix decomposition and deep learning, and reversely models the spectral mixing process. In addition, considering the sparsity and spatial smoothness of abundance, a fully-variation sparse constrained deep non-negative matrix factorization model for endmember matrix-oriented unmixing of hyperspectral remote sensing images is established. Through simulation experiments and real experiments, the proposed method is compared with five unmixing methods. The results show that compared with the abundance-oriented deep nonnegative matrix factorization algorithm, the average spectral angular distance and root mean square error of the proposed method are reduced, the best unmixing results are obtained.

Key words: hyperspectral remote sensing, hyperspectral unmixing, linear spectral unmixing, nonnegative matrix factorization, deep learning, deep nonnegative matrix factorization, sparsity constraint, total variation constraint

中图分类号:

TP751

赵文君,翟晗,张洪艳. 全变差稀疏约束深度非负矩阵分解高光谱遥感影像解混方法[J]. 电子科技, 2023, 36(2): 53-60.

ZHAO Wenjun,ZHAI Han,ZHANG Hongyan. Total Variation and Sparsity Regularized Deep Nonnegative Matrix Factorization for Hyperspectral Unmixing[J]. Electronic Science and Technology, 2023, 36(2): 53-60.

图/表 15

图1

图2

图3

图4

图5

图6

表1

图7

图8

图9

表2

图10

图11

图12

图13

参考文献 21

[1]	Zhai H, Zhang H Y, Zhang L P, et al. Sparsity-based clustering for large hyperspectral remote sensing images[J]. IEEE Transactions on Geoscience and Remote Sensing, 2020(11):1-15.
[2]	童庆禧, 张兵, 郑兰芬. 高光谱遥感:原理、技术与应用[M]. 北京: 高等教育出版社, 2006.
	Tong Qingxi, Zhang Bing, Zheng Lanfen. Hyperspectral remote sensing[M]. Beijing: Higher Education Press, 2006.
[3]	Keshava N, Mustard J F. Spectral unmixing[J]. IEEE Signal Processing Magazine, 2002, 19(1):44-57. doi: 10.1109/79.974727
[4]	Zhai H, Zhang H, Zhang L, et al. Laplacian-regularized low-rank subspace clustering for hyperspectral image band selection[J]. IEEE Transactions on Geoscience and Remote Sensing, 2018, 57(3):1723-1740. doi: 10.1109/TGRS.2018.2868796
[5]	Zhai H, Zhang H, Pingxiang L I, et al. Hyperspectral image clustering: Current achievements and future lines[J]. IEEE Geoscience and Remote Sensing Magazine, 2021, 17(7)1243-1247.
[6]	刘璐, 张洪艳, 张良培. 基于光谱加权低秩矩阵分解的高光谱影像去噪方法[J]. 电子科技, 2020, 33(5):21-27.
	Liu Lu, Zhang Hongyan, Zhang Liangpei. Hyperspectral image denoising via spectral weighted low-rank matrix approximation[J]. Electronic Science and Technology, 2020, 33(5):21-27.
[7]	魏一苇, 黄世奇. 高光谱混合像元分解技术研究综述[C]. 武汉: 第九届国家安全地球物理专题研讨会, 2013.
	Wei Yiwei, Huang Shiqi. Overview of the technique for hyperspectral unmixing[C]. Wuhan: The Ninth National Security Geophysics Symposium, 2013.
[8]	Jia S, Qian Y. Constrained nonnegative matrix factorization for hyperspectral unmixing[J]. IEEE Transactions on Geoscience and Remote Sensing, 2008, 47(1):161-173. doi: 10.1109/TGRS.2008.2002882
[9]	Lu X, Wu H, Yuan Y, et al. Manifold regularized sparse NMF for hyperspectral unmixing[J]. IEEE Transactions on Geoscience and Remote Sensing, 2012, 51(5):2815-2826. doi: 10.1109/TGRS.2012.2213825
[10]	Feng X R, Li H C, Li J, et al. Hyperspectral unmixing using sparsity-constrained deep nonnegative matrix factorization with total variation[J]. IEEE Transactions on Geoscience and Remote Sensing, 2018, 56(10):6245-6257. doi: 10.1109/TGRS.2018.2834567
[11]	Tong L, Zhou J, Qian B, et al. Adaptive graph regularized multilayer nonnegative matrix factorization for hyperspectral unmixing[J]. IEEE Journal of Selected Topics in Applied Earth Observations and Remote Sensing, 2020(13):434-447.
[12]	Lee D D, Seung H S. Learning the parts of objects by non-negative matrix factorization[J]. Nature, 1999, 401(6755):788-791. doi: 10.1038/44565
[13]	Qian Y, Jia S, Zhou J, et al. Hyperspectral unmixing via L1/2 sparsity-constrained nonnegative matrix factorization[J]. IEEE Transactions on Geoscience and Remote Sensing, 2011, 49(11):4282-4297. doi: 10.1109/TGRS.2011.2144605
[14]	He W, Zhang H, Zhang L. Total variation regularized reweighted sparse nonnegative matrix factorization for hyperspectral unmixing[J]. IEEE Transactions on Geoscience and Remote Sensing, 2017, 55(7):3909-3921. doi: 10.1109/TGRS.2017.2683719
[15]	Zhai H, Zhang H, Zhang L, et al. Total variation regularized collaborative representation clustering with a locally adaptive dictionary for hyperspectral imagery[J]. IEEE Transactions on Geoscience and Remote Sensing, 2018, 57(1):166-180. doi: 10.1109/TGRS.2018.2852708
[16]	Zhai H, Zhang H, Zhang L, et al. Nonlocal means regularized sketched reweighted sparse and low-rank subspace clustering for large hyperspectral images[J]. IEEE Transactions on Geoscience and Remote Sensing, 2020, 59(5):4164-4178. doi: 10.1109/TGRS.2020.3023418
[17]	Rudin L I, Osher S, Fatemi E. Nonlinear total variation based noise removal algorithms[J]. Physica D: Nonlinear Phenomena, 1992, 60(1-4):259-268. doi: 10.1016/0167-2789(92)90242-F
[18]	Nascimento J M P, Dias J M B. Vertex component analysis: A fast algorithm to unmix hyperspectral data[J]. IEEE Transactions on Geoscience and Remote Sensing, 2005, 43(4):898-910. doi: 10.1109/TGRS.2005.844293
[19]	Cichocki A, Zdunek R. Multilayer nonnegative matrix factorisation[J]. Electronics Letters, 2006, 42(16):947. doi: 10.1049/el:20060983
[20]	Boardman J W, Kruse F A, Green R O. Mapping target signatures via partial unmixing of AVIRIS data[J]. California Institute of Technology, 1995(3):1-4.
[21]	Zhu F, Wang Y, Fan B, et al. Spectral unmixing via data-guided sparsity[J]. IEEE Transactions on Image Processing, 2014, 23(12):5412-5427. pmid: 25330488

算法	平均光谱角距离值	均方根误差
VCA-FCLS	0.030 2	0.033 0
L₁_/₂-NMF	0.025 5	0.039 4
GNMF	0.024 4	0.032 7
MLNMF	0.059 6	0.038 0
SDNMF-TV	0.027 2	0.034 7
TVSDNMF	0.019 4	0.027 9

算法	平均光谱角距离值	均方根误差
VCA-FCLS	0.066 7	0.262 5
L_1/2-NMF	0.062 6	0.279 9
GNMF	0.065 4	0.264 3
MLNMF	0.061 9	0.267 4
SDNMF-TV	0.058 2	0.254 6
TVSDNMF	0.044 3	0.240 2