高光谱混合像元分解技术主要应用于处理遥感影像中存在的混合像元问题。然而，对于高光谱遥感影像而言，阴影问题同样也是一个非常严峻的挑战。高光谱遥感影像在获取的过程中，由于地物的遮挡等原因，阴影区域内的地物可能表现出与非阴影区域截然不同的光谱特征。这种由于阴影问题导致的地物光谱特征变化，使得大多数现有的混合像元分解方法难以精确地区分不同的地物，严重地影响了混合像元分解的准确性。因此，开发一种具有抗阴影影响的混合像元分解方法对于后续高光谱数据应用具有十分重要的意义。本文基于生成式对抗自编码网络开发了一种新的混合像元分解算法，显著削弱阴影问题对混合像元分解结果的影响，从而获取更加准确的混合像元分解结果。这一新方法称为“抗阴影混合像元分解”方法。本文主要研究内容如下：

（1）提出一种基于生成式对抗自编码网络抗阴影解混算法。该算法利用编码器中的光谱特征增强模块对阴影区域像元进行光谱特征增强，削弱了阴影区域对混合像元分解的影响。编码器使用空谱特征提取模块，充分利用了高光谱数据的空间和光谱信息并进行编码得到一组低维特征表示（丰度矩阵），并使用Softmax激活函数约束丰度的输出，使其满足丰度“和为1”以及非负约束。解码器实现高光谱影像盲源信号分离，获取纯净地物光谱（端元）。

（2）模拟阴影场景下算法验证。在模拟阴影场景下，GAA-AS算法与FCLS算法、AAS算法、DAEN算法、3DCNN算法通过使用50dB模拟数据进行测试。对比结果表明：GAA-AS算法能够正确地估计裸地（部分区域被阴影破坏）的丰度，其它解混算法处理阴影区域只能当成独立地物处理。其中，GAA-AS算法与地面真实丰度、真实地物端元之间对的差异最小，SAD（屋顶、裸地、植被）分别为：0.0148、0.0134、0.0115，Mean SAD为0.0112，RMSE为0.0219。

（3）真实阴影场景下与非阴影场景下算法验证。在真实阴影场景下，GAA-AS算法与FCLS算法、AAS算法、DAEN算法、3DCNN算法通过分别使用高光谱农业数据、高光谱林业数据分别进行测试。从定性层面对比分析，高光谱农业数据结果表明：GAA-AS算法能够正确估计地物端元裸地、植被（部分区域被屋顶、树木遮挡形成的阴影区域破坏）的丰度，而其它解混算法无法正确估计裸地、植被的丰度。同样地，高光谱林业数据对比结果表明：GAA-AS算法能够正确估计地物端元水泥地（部分区域被屋顶遮挡形成的阴影区域破坏）的丰度，而其它解混算法无法正确估计水泥地的丰度。

在真实非阴影场景下，GAA-AS算法与FCLS算法、AAS算法、DAEN算法、3DCNN算法通过分别使用Jasper数据、Samson数据分别进行测试。Jasper数据和Samson数据的对比结果表明，GAA-AS算法与其它解混算法相比，其解混精度会升高或保持稳定。以3DCNN算法作为基准，GAA-AS算法估计Jasper数据地物端元以及真实丰度的整体精度提高了0.93%、2.06%，GAA-AS算法估计Samson数据地物端元以及真实丰度的整体精度提高了2.6%、15.3%。

Hyperspectral unmixing can handle the issue of mixed pixels in a hyperspectral image. However, the problem of shadows presents a significant challenge for hyperspectral images as well. During the acquisition of hyperspectral images, due to object occlusions and other factors, the spectral features of objects in shadow areas may exhibit markedly different spectral features compared to non-shadow areas. The spectral variations of objects caused by shadow issues pose challenges to accurately distinguishing different objects for most existing spectral unmixing methods, seriously interfering with unmixing accuracy. Hence, developing an unmixing method with resistance to shadow effects is crucial for the subsequent applications of hyperspectral data. In this paper, a novel hyperspectral unmixing method is proposed based on Generative Adversarial Autoencoder (GAA), aiming to significantly mitigate the impact of shadow issues on the results of hyperspectral unmixing, thus achieving more accurate hyperspectral unmixing results. The newly developed method is defined as “anti-shadow unmixing”. The main research contents of this paper are as follows:

（1） A shadow-resistant unmixing algorithm based on generative adversarial autoencoder networks is proposed. This algorithm utilizes a spectral feature enhancement module in the encoder to enhance the spectral features of the pixels in shadow areas, thereby reducing the impact of shadow areas on hyperspectral unmixing. The encoder incorporates a spatial-spectral feature extraction module, fully leveraging the spatial and spectral information of hyperspectral data to encode and obtain a set of low-dimensional feature representations (abundance matrix). The output of abundances is constrained using a Softmax activation function, ensuring that the abundances satisfy the constraints of "sum to one" and non-negativity. The decoder achieves blind source separation of the hyperspectral image, obtaining the pure spectra (endmembers matrix).

（2） Validation of the algorithm under simulated shadow scenarios. In the simulated shadow scenarios, the GAA-AS algorithm was tested against the FCLS algorithm, AAS algorithm, DAEN algorithm, and 3DCNN algorithm using 50dB simulated data. The experimental results show that the GAA-AS algorithm can accurately estimate the abundance of bare ground (partially damaged by the shadow), whereas other unmixing algorithms can only treat shadow areas as independent land materials. Among them, the GAA-AS algorithm shows the smallest difference between the ground truth abundances and real land material endmembers. The Spectral Angle Distance (SAD) for roof, bare ground, and vegetation are respectively 0.0148, 0.0134, and 0.0115, with a Mean SAD of 0.0112 and RMSE of 0.0219.

（3） Validation of the algorithm in real shadow scenarios and  real non-shadow scenarios. In real shadow environments, the GAA-AS algorithm was tested against the FCLS algorithm, AAS algorithm, DAEN algorithm, and 3DCNN algorithm using both hyperspectral agricultural data and hyperspectral forestry data. Qualitatively, the results from the hyperspectral agricultural data indicate that the GAA-AS algorithm can accurately estimate the abundances of the endmembers such as bare ground and vegetation (partially damaged by shadows formed by the roof and vegetation), whereas other unmixing algorithms fail to accurately estimate the abundances of bare ground and vegetation. Similarly, the results from the hyperspectral forestry data show that the GAA-AS algorithm can accurately estimate the abundance of cement ground endmembers (partially damaged by shadows formed by the roof), while other unmixing algorithms fail to accurately estimate the abundance of cement ground.

In real non-shadow scenarios, the GAA-AS algorithm was tested against the FCLS algorithm, AAS algorithm, DAEN algorithm, and 3DCNN algorithm using Jasper data and Samson data respectively. The experimental results with Jasper and Samson data indicate that compared to other unmixing algorithms, the GAA-AS algorithm either improves or maintains stable unmixing accuracy. Using the 3DCNN algorithm as a benchmark, the GAA-AS algorithm increased the overall accuracy of estimating the  endmembers and real abundances in Jasper data by 0.93% and 2.06%, respectively. Similarly, for Samson data, the GAA-AS algorithm improved the overall accuracy of estimating the land material endmembers and real abundances by 2.6% and 15.3%, respectively.

[1]童庆禧, 张兵, 张立福. 中国高光谱遥感的前沿进展[J]. 遥感学报, 2016, 20(5):689-707.

[2]张淳民, 穆廷魁, 颜廷昱等. 高光谱遥感技术发展与展望[J]. 航天返回与遥感, 2018, 39(3): 104-114.

[3]Peyghambari S, Zhang Y. Hyperspectral remote sensing in lithological mapping, mineral exploration, and environmental geology: an updated review[J]. Journal of Applied Remote Sensing, 2021, 15(3): 031501-031501.

[4]张兵. 高光谱影像处理与信息提取前沿[J]. 遥感学报, 2016, 20(5): 1062-1090.

[5]Zhu D, Du B, Zhang L. Single-spectrum-driven binary-class sparse representation target detector for hyperspectral imagery[J]. IEEE Transactions on Geoscience and Remote Sensing, 2020, 59(2): 1487-1500.

[6]Moharram M A, Sundaram D M. Dimensionality reduction strategies for land use land cover classification based on airborne hyperspectral imagery: a survey[J]. Environmental Science and Pollution Research, 2023, 30(3): 5580-5602.

[7]Ding X, Hu W. Advancements in Geological Disaster Monitoring and Early Warning Systems: A Deep Learning and Computer Vision Approach[J]. Traitement du Signal, 2023, 40(3): 1195.

[8]Keshava N, Mustard J F. Spectral unmixing[J]. IEEE signal processing magazine, 2002, 19(1): 44-57.

[9]Iordache M D, Bioucas-Dias J M, Plaza A. Sparse unmixing of hyperspectral data[J]. IEEE Transactions on Geoscience and Remote Sensing, 2011, 49(6): 2014-2039.

[10]Rosario-Torres S, Vélez-Reyes M. An algorithm for fully constrained abundance estimation in hyperspectral unmixing[C]//Algorithms and Technologies for Multispectral, Hyperspectral, and Ultraspectral Imagery XI. SPIE, 2005, 5806: 711-719.

[11]Guo A J X, Zhu F. Improving deep hyperspectral image classification performance with spectral unmixing[J]. Signal Processing, 2021, 183: 107949.

[12]李晓松, 高志海, 李增元等. 基于高光谱混合像元分解的干旱地区稀疏植被覆盖度估测[J]. 应用生态学报, 2010, 21(1): 152-158.

[13]崔宾阁, 庄仲杰, 任广波等. 典型高光谱影像端元提取算法在黄河口湿地应用评价研究[J]. 海洋科学, 2015,39(2): 104-109.

[14]田原. 基于边缘检测和噪声去除的遥感阴影检测算法[J]. 电子技术与软件工程, 2015(23): 98-99.

[15]Silva G F, Carneiro G B, Doth R, et al. Near real-time shadow detection and removal in aerial motion imagery application[J]. ISPRS Journal of photogrammetry and remote sensing, 2018, 140: 104-121.

[16]Mostafa Y, Abdelhafiz A. Accurate shadow detection from high-resolution satellite images[J]. IEEE Geoscience and Remote Sensing Letters, 2017, 14(4): 494-498.

[17]Dan L I, Baosheng W U, Bowei C, et al. Review of water body information extraction based on satellite remote sensing[J]. Journal of Tsinghua University (Science and Technology), 2020, 60(2): 147-161.

[18]Yu H, Chi J, Shang X, et al. Robust linear unmixing with enhanced constraint of classification for hyperspectral remote sensing imagery[J]. IET Image Processing, 2022, 16(13): 3557-3566.

[19]丰远远, 高贤君, 杨元维等. 改进对数变换联合局部增强的高分遥感影像阴影自动补偿方法[J]. 激光与光电子学进展, 2020, 57(20): 67-75.

[20]Manolakis D, Siracusa C, Shaw G. Hyperspectral subpixel target detection using the linear mixing model[J]. IEEE transactions on geoscience and remote sensing, 2001, 39(7): 1392-1409.

[21]Altmann Y, Dobigeon N, McLaughlin S, et al. Nonlinear spectral unmixing of hyperspectral images using Gaussian processes[J]. IEEE Transactions on Signal Processing, 2013, 61(1): 2442-2453.

[22]Winter M E. N-FINDR: An algorithm for fast autonomous spectral endmember determination in hyperspectral data[C]. in Proc. SPIE. Imaging Spectrometry V. 1999, 3753: 266-275.

[23]Fuhrmann D R. Simplex shrink-wrap algorithm[C]. Automatic Target Recognition IX. SPIE, 1999, 3718: 501-511.

[24]Plaza A, Martinez P, Pérez R, et al. Spatial/spectral endmember extraction by multidimensional morphological operations[J]. IEEE transactions on geoscience and remote sensing, 2002, 40(9): 2025-2041.

[25]Mei S, He M, Wang Z, et al. Spatial purity based endmember extraction for spectral mixture analysis[J]. IEEE Transactions on Geoscience and Remote Sensing, 2010, 48(9): 3434-3445.

[26]Li J, Agathos A, Zaharie D, et al. Minimum volume simplex analysis: A fast algorithm for linear hyperspectral unmixing[J]. IEEE Transactions on Geoscience and Remote Sensing, 2015, 53(9): 5067-5082.

[27]Lee D, Seung H S. Learning the parts of objects by non-negative matrix factorization[J]. Nature, 1999, 401(6755): 788-791.

[28]Miao L, Qi H. Endmember extraction from highly mixed data using minimum volume constrained nonnegative matrix factorization[J]. IEEE Transactions on Geoscience and Remote Sensing, 2007, 45(3): 765-777.

[29]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.

[30]Guo R, Wang W, Qi H. Hyperspectral image unmixing using autoencoder cascade[C]. 2015 7th Workshop on Hyperspectral Image and Signal Processing: Evolution in Remote Sensing （WHISPERS）. IEEE, 2015: 1-4.

[31]Su Y, Marinoni A, Li J, et al. Stacked nonnegative sparse autoencoders for robust hyperspectral unmixing[J]. IEEE Geoscience and Remote Sensing Letters, 2018, 15(9): 1427-1431.

[32]Palsson B, Sigurdsson J, Sveinsson J R, et al. Hyperspectral unmixing using a neural network autoencoder[J]. IEEE Access, 2018, 6: 25646-25656.

[33]Qu Y, Qi H. uDAS: An untied denoising autoencoder with sparsity for spectral unmixing[J]. IEEE Transactions on Geoscience and Remote Sensing, 2018, 57(3): 1698-1712.

[34]Su Y, Li J, Plaza A, et al. DAEN: Deep autoencoder networks for hyperspectral unmixing[J]. IEEE Transactions on Geoscience and Remote Sensing, 2019, 57(7): 4309-4321.

[35]Jin Q, Ma Y, Fan F, et al. Adversarial autoencoder network for hyperspectral unmixing[J]. IEEE Transactions on Neural Networks and Learning Systems, 2021, 34(8): 4555-4569.

[36]Su Y, Xu X, Li J, et al. Deep autoencoders with multitask learning for bilinear hyperspectral unmixing[J]. IEEE Transactions on Geoscience and Remote Sensing, 2020, 59(10): 8615-8629.

[37]Hong D, Gao L, Yao J, et al. Endmember-guided unmixing network (EGU-Net): A general deep learning framework for self-supervised hyperspectral unmixing[J]. IEEE Transactions on Neural Networks and Learning Systems, 2021, 33(11): 6518-6531.

[38]Gao L, Han Z, Hong D, et al. CyCU-Net: Cycle-consistency unmixing network by learning cascaded autoencoders[J]. IEEE Transactions on Geoscience and Remote Sensing, 2021, 60: 1-14.

[39]Hua Z, Li X, Qiu Q, et al. Autoencoder network for hyperspectral unmixing with adaptive abundance smoothing[J]. IEEE Geoscience and Remote Sensing Letters, 2020, 18(9): 1640-1644.

[40]Zhang G, Mei S, Xie B, et al. Spectral variability augmented sparse unmixing of hyperspectral images[J]. IEEE Transactions on Geoscience and Remote Sensing, 2022, 60: 1-13.

[41]Yang X, Chen J, Wang C, et al. Residual dense autoencoder network for nonlinear hyperspectral unmixing[J]. IEEE Journal of Selected Topics in Applied Earth Observations and Remote Sensing, 2022, 15: 5580-5595.

[42]Li Y, Jiang B, Li X, et al. Unsupervised hyperspectral unmixing based on robust nonnegative dictionary learning[J]. Journal of Systems Engineering and Electronics, 2022, 33(2): 294-304.

[43]Palsson B, Ulfarsson M O, Sveinsson J R. Convolutional autoencoder for spatial-spectral hyperspectral unmixing[C]//IGARSS 2019-2019 IEEE International Geoscience and Remote Sensing Symposium. IEEE, 2019: 357-360.

[44]Zhao M, Shi S, Chen J, et al. A 3-D-CNN framework for hyperspectral unmixing with spectral variability[J]. IEEE Transactions on Geoscience and Remote Sensing, 2022, 60: 1-14.

[45]Qi L, Gao F, Dong J, et al. SSCU-Net: Spatial-spectral collaborative unmixing network for hyperspectral images[J]. IEEE Transactions on Geoscience and Remote Sensing, 2022, 60: 1-15.

[46]Vargas M, Toral S L, Milla J M, et al. A shadow removal algorithm for vehicle detection based on reflectance ratio and edge density[C]//13th International IEEE Conference on Intelligent Transportation Systems. IEEE, 2010: 1123-1128.

[47]Liu X, Wang C, Meng Y, et al. Target detection based on spectral derivation in HSI shadow region classified by convolutional neural networks[J]. Canadian Journal of Remote Sensing, 2019, 45(6): 782-794.

[48]Karim S, Zhang Y, Ali S, et al. An improvement of vehicle detection under shadow regions in satellite imagery[C]//Ninth International Conference on Graphic and Image Processing (ICGIP 2017). SPIE, 2018, 10615: 1243-1249.

[49]Adler-Golden S M, Matthew M W, Anderson G P, et al. Algorithm for de-shadowing spectral imagery[C]//Imaging Spectrometry VIII. SPIE, 2002, 4816: 203-210.

[50]Richter R, Müller A. De-shadowing of satellite/airborne imagery[J]. International Journal of Remote Sensing, 2005, 26(15): 3137-3148.

[51]Ramakrishnan R, Nieto J, Scheding S. Shadow compensation for outdoor percep-tion[C]//2015 IEEE International Conference on Robotics and Automation (ICRA). IEEE, 2015: 4835-4842.

[52]Tolt G, Shimoni M, Ahlberg J. A shadow detection method for remote sensing images using VHR hyperspectral and LIDAR data[C]//2011 IEEE international geoscience and remote sensing symposium. IEEE, 2011: 4423-4426.

[53]Zosso D, Tran G, Osher S. A unifying retinex model based on non-local differential operators[C]//Computational Imaging XI. SPIE, 2013, 8657: 865702.

[54]Liu Y, Bioucas-Dias J, Li J, et al. Hyperspectral cloud shadow removal based on linear unmixing[C]//2017 IEEE International Geoscience and Remote Sensing Symposium (IGARSS). IEEE, 2017: 1000-1003.

[55]Wang Q, Yuan Z, Du Q, et al. GETNET: A general end-to-end 2-D CNN framework for hyperspectral image change detection[J]. IEEE Transactions on Geoscience and Remote Sensing, 2018, 57(1): 3-13.

[56]Windrim L, Ramakrishnan R, Melkumyan A, et al. A physics-based deep learning approach to shadow invariant representations of hyperspectral images[J]. IEEE Transactions on Image Processing, 2017, 27(2): 665-677.

[57]Zhao M, Yan L, Chen J. Hyperspectral image shadow compensation via cycle-consistent adversarial networks[J]. Neurocomputing, 2021, 450: 61-69.

[58]张帅洋, 华文深, 应家驹等. 高光谱线性解混研究进展[J]. 激光杂志, 2021, 42(3): 17-21.

[59]Khajehrayeni F, Ghassemian H. A linear hyperspectral unmixing method by means of autoencoder networks[J]. International Journal of Remote Sensing, 2021, 42(7): 2517-2531.

[60]杨斌, 王斌. 高光谱遥感图像非线性解混研究综述[J]. 红外与毫米波学报, 2017, 36(2): 173-185.

[61]苏远超, 许若晴, 高连如等. 基于深度学习的高光谱遥感图像混合像元分解研究综述[J]. 遥感学报, 2024, 28(1): 1-19.

[62]Ma W, Gong C, Hu Y, et al. The Hughes phenomenon in hyperspectral classification based on the ground spectrum of grasslands in the region around Qinghai Lake[C]//International Symposium on Photoelectronic Detection and Imaging 2013: Imaging Spectrometer Technologies and Applications. SPIE, 2013, 8910: 363-373.

[63]Lalitha V, Latha B. A novel model for hyper spectral image enhancement and classification: PCA, MBAO and CNN integration[J]. Optical and Quantum Electronics, 2024, 56(3): 473.

[64]Nielsen A A. Kernel maximum autocorrelation factor and minimum noise fraction transformations[J]. IEEE Transactions on Image Processing, 2010, 20(3): 612-624.

[65]Abdi H, Williams L J. Principal component analysis[J]. Wiley interdisciplinary reviews: computational statistics, 2010, 2(4): 433-459.

[66]Boardman J W. Automating spectral unmixing of AVIRIS data using convex geometry concepts[C]//JPL, Summaries of the 4th Annual JPL Airborne Geoscience Workshop. Volume 1: AVIRIS Workshop. 1993. 1(1).11-14.

[67]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.

[68]Xiong W, Chang C I, Wu C C, et al. Fast algorithms to implement N-FINDR for hyperspectral endmember extraction[J]. IEEE Journal of selected topics in applied earth observations and remote sensing, 2011, 4(3): 545-564.

[69]Harsanyi J C, Chang C I. Hyperspectral image classification and dimensionality reduction: An orthogonal subspace projection approach[J]. IEEE Transactions on geoscience and remote sensing, 1994, 32(4): 779-785.

[70]Berman M, Kiiveri H, Lagerstrom R, et al. ICE: A statistical approach to identifying endmembers in hyperspectral images[J]. IEEE transactions on Geoscience and Remote Sensing, 2004, 42(10): 2085-2095.

[71]Suykens J A K, Vandewalle J. Least squares support vector machine classifiers[J]. Neural processing letters, 1999, 9: 293-300.

[72]Ashton E A, Schaum A. Algorithms for the detection of sub-pixel targets in multispectral imagery[J]. Photogrammetric Engineering & Remote Sensing, 1998, 64(7): 723-731.

[73]Heinz D C. Fully constrained least squares linear spectral mixture analysis method for material quantification in hyperspectral imagery[J]. IEEE transactions on geoscience and remote sensing, 2001, 39(3): 529-545.

[74]Settle J J, Drake N A. Linear mixing and the estimation of ground cover proportions[J]. International Journal of Remote Sensing, 1993, 14(6): 1159-1177.

[75]Bro R, De Jong S. A fast non-negativity-constrained least squares algorithm[J]. Journal of Chemometrics: A Journal of the Chemometrics Society, 1997, 11(5): 393-401.

[76]Meng Q, Catchpoole D, Skillicom D, et al. Relational autoencoder for feature extraction[C]//2017 International joint conference on neural networks (IJCNN). IEEE, 2017: 364-371.

[77]Zhang G, Liu Y, Jin X. A survey of autoencoder-based recommender systems[J]. Frontiers of Computer Science, 2020, 14: 430-450.

[78]Zhao R, Zhan M, Deng X, et al. Yet another traffic classifier: A masked autoencoder based traffic transformer with multi-level flow representation[C]//Proceedings of the AAAI Conference on Artificial Intelligence. 2023, 37(4): 5420-5427.

[79]宋尚真, 杨怡欣, 王会峰等. 高光谱影像稀疏约束与自编码器特征提取相结合的异常检测方法[J]. 测绘学报, 2023, 52(6): 932-943.

[80]Özdemir O B, Koz A. 3D-CNN and Autoencoder-based Gas Detection in Hyperspectral Images[J]. IEEE Journal of Selected Topics in Applied Earth Observations and Remote Sensing, 2023, 16: 1474-1482.

[81]Goodfellow I, Pouget-Abadie J, Mirza M, et al. Generative adversarial networks[J]. Communications of the ACM, 2020, 63(11): 139-144.

[82]Paladugu P S, Ong J, Nelson N, et al. Generative adversarial networks in medicine: important considerations for this emerging innovation in artificial intelligence[J]. Annals of biomedical engineering, 2023, 51(10): 2130-2142.

[83]Vallabhaneni R, Vaddadi S A, Pillai S, et al. Detection of cyberattacks using bidirectional generative adversarial network[J]. Indonesian Journal of Electrical Engineering and Computer Science, 2024, 35(3): 1653-1660.

[84]Zhu J Y, Park T, Isola P, et al. Unpaired image-to-image translation using cycle-consistent adversarial networks[C]//Proceedings of the IEEE international conference on computer vision. 2017: 2223-2232.

[85]Wang X, Li Y, Cheng Y. Hyperspectral image classification based on unsupervised heterogeneous domain adaptation cyclegan[J]. Chinese Journal of Electronics, 2020, 29(4): 608-614.

[86]Hossain M A, Sajib M S A. Classification of image using convolutional neural network (CNN)[J]. Global Journal of Computer Science and Technology, 2019, 19(D2): 13-18.

[87]Lu Y, Tao X, Jiang F, et al. Image recognition of rice leaf diseases using atrous convolutional neural network and improved transfer learning algorithm[J]. Multimedia Tools and Applications, 2024, 83(5): 12799-12817.

[88]Xu S, Geng S, Fan T, et al. Hyperspectral image classification method based on narrowing semantic gap convolutional neural network[J]. International Journal of Remote Sensing, 2024, 45(7): 2208-2234.

[89]Chen Y, Zhang H, Liu L, et al. Research on image inpainting algorithm of improved total variation minimization method[J]. Journal of Ambient Intelligence and Humanized Computing, 2023, 14(5): 5555-5564.

[90]Ambrosio L, Aziznejad S, Brena C, et al. Linear inverse problems with Hessian–Schatten total variation[J]. Calculus of Variations and Partial Differential Equations, 2024, 63(1): 9.

[91]Zhu H, Luo J, He S. Detecting Multiple Mixed Bacteria Using Dual-Mode Hyperspectral Imaging and Deep Neural Networks[J]. Applied Sciences, 2024, 14(4): 1525.

[92]Toh S C, Lai S H, Mirzaei M, et al. Sequential Data Processing for IMERG Satellite Rainfall Comparison and Improvement Using LSTM and ADAM Optimizer[J]. Applied Sciences, 2023, 13(12): 7237.

[93]Wang M, Zhao M, Chen J, et al. Nonlinear unmixing of hyperspectral data via deep autoencoder networks[J]. IEEE Geoscience and Remote Sensing Letters, 2019, 16(9): 1467-1471.

[94]Chang C I, Wu C C, Liu W, et al. A new growing method for simplex-based endmember extraction algorithm[J]. IEEE transactions on geoscience and remote sensing, 2006, 44(10): 2804-2819.