Abstract：In recent years，with the rapid development of military reconnaissance and identification technology，military equipment used for reconnaissance and detection has gradually achieved high-precision levels. The troops with high-tech reconnaissance methods can often perform precise strikes on targets, significantly reducing the cost of victory in war. The more mature hyperspectral imaging methods include satellite remote sensing and high-altitude aerial imaging technologies. The two imaging methods have roughly the same reconnaissance time and the same direction of incident light. Therefore, the spectral curve of the ground object is relatively fixed. However, under land-based conditions, the spectral curve of the ground feature is prominently affected by the imaging environment, so the method of hyperspectral image classification is suitable for land-based conditions should be studied. In land-based hyperspectral images, the identification and classification of each feature are beneficial to the subsequent identification and processing of camouflage targets. Different from traditional remote sensing spectral image classification, the classification of hyperspectral camouflage targets under land-based conditions is not only difficult to obtain training samples, and in hyperspectral images under land-based conditions, the correlation between training samples under land-based conditions, the correlation between training samples varies with the target type. The parameters of the detector and the imaging environment are constantly changing. Classification methods based on sparse representation have been widely used to deal with image problems and various machine vision problems, including hyperspectral image classification. For land-based hyperspectral images, sparse coding strategies based on fixed norm constraints cannot be adapted under land-based conditions, hyperspectral imaging. For land-based hyperspectral images, sparse coding strategies based on fixed norm constraints cannot be adapted. Under land-based conditions, hyperspectral imaging is a changeable environment, and adaptive sparse representation can adaptively adjust norm constraints based on sample correlation. Correlation coefficients can improve the image’s recognition accuracy of destructive factors (shadows, noise points, etc.). This paper proposes a new hyperspectral image classification method by introducing regularization parameters, fusing adaptive sparse representation and correlation coefficients. In order to verify the effectiveness of the proposed method, camouflage objects were set in the green vegetation background and the desert background, and different classification methods classified the images. The experimental results show that the method in this paper is obvious, whether it is classification accuracy or classification consistency. The advantages of this can be applied to the classification of hyperspectral images under land-based conditions, providing a theoretical basis for camouflage reconnaissance and identification.
周 冰，李秉璇，贺 宣，刘贺雄，王法臻. 融合自适应稀疏表示和相关系数的高光谱伪装分类方法[J]. 光谱学与光谱分析, 2021, 41(12): 3851-3856.
ZHOU Bing, LI Bing-xuan, HE Xuan, LIU He-xiong，WANG Fa-zhen. Classification of Camouflages Using Hyperspectral Images Combined With Fusing Adaptive Sparse Representation and Correlation Coefficient. SPECTROSCOPY AND SPECTRAL ANALYSIS, 2021, 41(12): 3851-3856.
[1] PENG Yao-qi, XIAO Ying-xin, ZHENG Yong-jun, et al(彭要奇, 肖颖欣, 郑永军, 等). Spectroscopy and Spectral Analysis(光谱学与光谱分析), 2020, 40(5): 1356.
[2] Zhang Q, Li B. Discriminative K-SVD for Dictionary Learning in Face Recognition, 2010 IEEE Computer Society Conference on Computer Vision and Pattern Recognition, 2010, doi: 10.1109/CVPR.2010.5539989.
[3] HUANG Chuan-lu, CHAO Kun, MAO Yun-zi(黄传禄，晁 坤，毛云志). Chinese Journal of Radio Science(电波科学学报)，2014，29(1)：150.
[4] Zhu T, Bao W, Shen X, et al. Journal of Applied Remote Sensing, 2020, 14(2): 1.
[5] Akila D, Bhaumik A, Doss S, et al. Hyperspectral Image Classification by Means of Suprepixel Representation with KNN. in: Intelligent Computing and Innovation on Data Science，Lecture Notes in Network and Systems Vol 118. Springer, 2020, https://doi.org/10.1007/978-981-15-3284-9_42.
[6] Yang W, Peng J, Sun W, et al. IEEE Journal of Selected Topics in Applied Earth Observations and Remote Sensing, 2020, 12(12): 5023.
