It is critical to decompose mixed pixels in ahyperspectral image (HSI)into pure spectral signatures and fractions, known as endmembers and abundances. However, current methods usually combine all endmembers and abundances into a single matrix. However, this approach overlooks the distinct capacity differences of each substance subspace. Furthermore, traditional approaches typically reconstruct without accounting for corresponding errors, resulting in suboptimal outcomes. In this work, we introduce a novel framework that uses self-adjustedindividual logical feature (LIF)subspaces for each substance. This enables the accurate modeling of each substance’s unique properties. Our method calculates the capacity of each subspace by unifying the features of each substance, thereby ensuring a more accurate representation. Importantly, our approach balances reconstruction fidelity and error, preventing blind approximation of the observed HSI and addressing overfitting. Additionally, our approach exploits correlations between reconstructed subspaces to minimize redundancy. Extensive experimental results on several datasets demonstrate the superior performance and validity of the proposed method.

There are multiple challenges in small object detection, including limited instances, insufficient features, diverse scales, uneven distribution, ambiguous boundaries, and complex backgrounds. These issues often lead to high false detection rates and hinder model generalization and convergence. This study proposes a multi-scale object detection algorithm that enhances the detection of subtle features by improving the detection head and incorporating a minimum point distance intersection-over-union loss. The enhanced detection head improves target representation, enabling more precise localization and classification of small objects. Meanwhile, the new loss function stabilizes bounding box regression by adaptively adjusting auxiliary bounding box scales. Evaluations on two benchmark datasets demonstrate that our method achieves a 2.6% increase in mAP50 and a 1.8% improvement in mAP50:95 on the Satellite Imagery Multivehicles dataset and a 1.9% increase in mAP50:95 on the DIOR dataset. Furthermore, the model reduces the number of parameters by 2.5% and the computational cost by 1.4%, demonstrating its potential for real-time detection applications.

Convolutional neural network (CNN)is used broadly forHyperspectral image classification (HIC). However, the existing CNN-based methods usually oversimplify even distort the factual relationships between underlying characteristics in actual scenes. To this end, we employ theoblique rotation factor (ORF)analyses to build a robust framework for few-shot HIC. Under the guidance of ORF, double-branch 3-dimensional (D)CNNs are restructured via discarding all the batch normalization units to model the factual relationship among the latent characteristics, and two separate progressive attention schemes are embedded respectively in the double-branch 3DCNNs to exploit the spatial/spectral details fully. To alleviate the issue of model degradation, the double-branch 3DCNNs are formed with the residual connections. Then to refine the features further and improve the efficiency by preventing overfitting, we restructure 2DCNN by employing the dropout to replace the pooling layers. Finally, the restructured double-branch 3DCNNs and 2DCNN are assembled to form an effective framework for the task of fewshot HIC. The proposed framework is validated on the datasets Loukia, Trento, andKennedy Space Center (KSC), respectively 6%, 1%, and 6% samples are employed to train, the overall accuracy still achieves 88.84%, 97.8% and 99.39%. It indicates that the constructed framework has advantages in few-shot learning.

Factor analysis (FA)is commonly used in fields such as economics and now being introduced as a new tool ondimensionality reduction (DR)forhyperspectral image classification (HSIC), but FA usually employed orthogonal rotation to directly maximize the separation among factors, which would oversimplify the relationships between variables and factors, worse still, the orthogonal rotation often distorts the true relationships between underlying traits in real life and can not always accurately represent these relationships. To this end, this letter proposes a DR algorithm about FA based on oblique rotation Oblimax to improve HSIC. Firstly, the common factors are extracted from the hyperspectral data to form a factor loading matrix which will be obliquely rotated, then its factor score is estimated to obtain the eigen dimensions for the hyperspectral data, thus realizing DR. On the basis of the successful DR, a deep classifier is constructed, specially, a double-branch structure about3 dimensional-convolutional neural networks (3D-CNN)with different sizes is restructured to extract multi-scale spatial-spectral features, and early fusion is performed on the features, then2 dimensional-convolutional neural networks (2D-CNN)is restructured to reduce the computational complexity and learns more spatial features. Finally, the accuracy of the proposed algorithm on the datasets Indian Pines, Kennedy Space Center and Muufl Gulfport, respectively achieves 99.78%, 99.95% and 95.57%. It shows that the proposed algorithm in this letter has advantages in improving the classification accuracy and reducing the complexity of computation.

Blind hyperspectral unmixing (BHU) is an important technology to decompose the mixed hyperspectral image (HSI), which is actually an ill-posed problem. The ill-posedness of the BHU is deteriorated by nonlinearity, endmember variability (EV) and abnormal points, which are considered as three challenging intractable interferences currently. To sidestep the challenges, we present a novel unmixing model, where a latent multidiscriminative subspace is explored and the inherent self-expressiveness property is employed. The most existing unmixing approaches directly decompose the HSI utilizing original features in an interference corrupted single subspace, unlike them, our model seeks the underlying intrinsic representation and simultaneously reconstructs HSI based on the learned latent subspace. With the help of both clustering homogeneity and intrinsic features selection, structural differences in the HSI and the spectral property of a certain material are exploited perfectly, and an ideal multiheterogeneous subspace is recovered from the heavily contaminated original HSI. Based on the multiheterogeneous subspace, the reconstructed differentiated transition matrix is split into two matrices to avoid the emergence of the artificial endmember. Experiments are conducted on synthetic and four representative real HSI sets, and all the experimental results demonstrate the validity and superiority of our proposed method.

Blind hyperspectral unmixing (BHSU) is ill-posedness. It aims to obtain accurate and robust endmember signatures and the corresponding abundances simultaneously. Nonnegative matrix factorization (NMF)-based sparsity-regularized algorithms have been widely employed for the BHSU. However, the existing unmixing approaches are sensitive to the multifarious intrinsic interferences and noises, which are caused because of the utilization of the inappropriate loss function to measure the quality of the hyperspectral data (HD) reconstruction and regularization. In this article, we propose a noise-free graph regularized model (NFGRM) by applying the dual graph regularized robust nonnegative matrix tri-factorization (NMTF), which leads to a novel reliable reconstruction of the HD. In the NFGRM, all the challenging interferences are addressed as noises. Consequently, a more faithful approximation is expected to recover from the highly noisy mixed data set and achieve robust regularization by controlling the heteroscedastic noises and the ill-posedness of the BHSU problem simultaneously. Experimental results on synthetic and several benchmark HD sets demonstrate the effectiveness and robustness of the proposed model and algorithm.

In this letter, a sparse-constrained hyperspectral unmixing method via reconstruction error approximation is proposed. In the presented method, all the noises and the outliers are treated as diverse interferences and addressed to minimize the regularization error. Several techniques are involved in our presented approach: 1) to attenuate the interference of noise, an auxiliary variable is introduced; 2) based on the relative noiseless hyperspectral image, a sparse constraint is employed to achieve the sparsity; and 3) besides, the correntropy-induced metric (CIM), instead of the$L_{2}$- or$L_{2,1}$-norm loss function, is utilized to measure the quality of the unmixing model approximation. A series of experiments on the synthetic and real hyperspectral images is conducted, and all the experiment results show the efficacy of the proposed approach.

Traditional spectral unmixing methods are usually based on the linear mixture model (LMM) or nonlinear mixture model (NLMM), in which only the additive noise is considered. However, in hyperspectral applications, the additive, multiplicative, and mixed noises play important roles. In this paper, we propose an antinoise model for hyperspectral unmixing. In the antinoise model, all the additive, multiplicative and mixed noises are addressed. To deal with the problems faced by LMM or NLMM and to tackle the antinoise model, an antinoise model based hyperspectral unmixing method is presented, where block coordinate descent is employed to solve an approximatedL0norm constraint, then a nonnegative matrix factorization (NMF) method is presented, which is based on the bounded Itakura-Saito divergence. The experimental results on both synthetic and real hyperspectral data sets demonstrate the efficacy of the proposed model and the corresponding method.

Multicommodity flow-based virtual network embedding algorithm does not consider link energy, which causes waste of energy. And its high time complexity can not meet real-time requirements of online virtual network embedding. In this paper, we find the dynamic inversion phenomenon, where revenue does not rely on embedding cost. Two novel link mapping algorithms are proposed for path splitting which based on the undirected network minimum cost flow. They enable link resource to consolidate and have low time complexity. Simulation results show that proposed algorithms reduce energy consumption and ensure real-time performance of online VN embedding.