In recent years, sparse unmixing has attracted significant attention as it can effectively avoid the bottleneck problems associated with the absence of pure pixels and the estimation of the number of endmembers in hyperspectral scenes. The joint-sparsity model has outperformed the single sparse unmixing method. However, the joint-sparsity model might cause some aliasing artifacts for the pixels on the boundaries of different constituent endmembers. To address this shortcoming, researchers have developed many unmixing algorithms based on low-rank representation, which makes good use of the global structure of data. In addition, the high mutual coherence of spectral libraries strongly affects the applicability of spare unmixing. In this study, adopting combined constraints imposing sparsity and low rankness, a novel algorithm called non-convex joint-sparsity and low-rank unmixing with dictionary pruning (NCJSpLRUDP) is developed. In particular, we impose sparsity on the abundance matrix using the l2,p mixed norm, and we also employ the weighted Schatten p-norm instead of the convex nuclear norm as an approximation for the rank. The key parameter p is set between 0.4 and 0.6, and a good quality sparse solution is generated. The effectiveness of the proposed algorithm is demonstrated on both simulated and real hyperspectral datasets.
How to extract the indicative signatures from the spectral data is an important issue for further retrieval based on remote sensing technique. This study provides new insight into extracting indicative signatures by identifying oblique extremum points, rather than local extremum points traditionally known as absorption points. A case study on retrieving soil organic matter (SOM) contents from the black soil region in Northeast China using spectral data revealed that the oblique extremum method can effectively identify weak absorption signatures hidden in the spectral data. Moreover, the comparison of retrieval outcomes using various indicative signature extraction methods reveals that the oblique extremum method outperforms the correlation analysis and traditional extremum methods. The experimental findings demonstrate that the radial basis function (RBF) neural network retrieval model exposes the nonlinear relationship between reflectance (or reflectance transformation results) and the SOM contents. Additionally, an improved oblique extremum method based on the second-order derivative is provided. Overall, this research presents a novel perspective on indicative signature extraction, which could potentially offer better retrieval performance than traditional methods.
1 September 2019 in IEICE Transactions on Information and Systems
The sparse unmixing of hyperspectral data has attracted much attention in recent years because it does not need to estimate the number of endmembers nor consider the lack of pure pixels in a given hyperspectral scene. However, the high mutual coherence of spectral libraries strongly affects the practicality of sparse unmixing. The collaborative sparse unmixing via variable splitting and augmented Lagrangian (CLSUnSAL) algorithm is a classic sparse unmixing algorithm that performs better than other sparse unmixing methods. In this paper, we propose a CLSUnSAL-based hyperspectral unmixing method based on dictionary pruning and reweighted sparse regression. First, the algorithm identifies a subset of the original library elements using a dictionary pruning strategy. Second, we present a weighted sparse regression algorithm based on CLSUnSAL to further enhance the sparsity of endmember spectra in a given library. Third, we apply the weighted sparse regression algorithm on the pruned spectral library. The effectiveness of the proposed algorithm is demonstrated on both simulated and real hyperspectral datasets. For simulated data cubes (DC1, DC2 and DC3), the number of the pruned spectral library elements is reduced by at least 94% and the runtime of the proposed algorithm is less than 10% of that of CLSUnSAL. For simulated DC4 and DC5, the runtime of the proposed algorithm is less than 15% of that of CLSUnSAL. For the real hyperspectral datasets, the pruned spectral library successfully reduces the original dictionary size by 76% and the runtime of the proposed algorithm is 11.21% of that of CLSUnSAL. These experimental results show that our proposed algorithm not only substantially improves the accuracy of unmixing solutions but is also much faster than some other state-of-the-art sparse unmixing algorithms.
22 January 2025 in IEEE Journal of Selected Topics in Applied Earth Observations and Remote Sensing
Identification of spectrally similar materials from multispectral remote sensing (RS) imagery with only several bands is an important issue that challenges comprehensive applications of the RS of surface characteristics. This study proposes a new method to identify spectrally similar materials from these types of imagery. The method is constructed based on the theory of condition number of matrix, and a theorem is proven as the foundation of the designed identification algorithm. Mathematically, the motivation behind designing this new algorithm is to decrease the condition number of the matrix for a linear system and, by doing so, to change an ill-conditioned system to a well-conditioned one. Technically, this new method achieves the purpose by adding supplementary features to all the original spectra including similar materials, which can be further used as indicative signatures to identify these materials. Thus, the proposed method is named a Condition Numberbased Method with Supplementary Features (SF-CNM). The threshold scheme and supplementary features are two main novelty techniques to ensure the uniqueness and accuracy of the proposed SF-CNM for specified samples. The results for a case study to identify water, ice, snow, shadow, and other materials from Landsat 8 OLI data indicate that SF-CNM can identify the materials specified by the given samples successfully and accurately and that SF-CNM significantly outperforms those of Spectral Angle Mapper algorithm (SAM), Mahalanobis Classifier (MC), Maximum Likelihood (ML) and Artificial Neural Network (ANN), and produces the performance similar to, even slightly better than that of Support Vector Machine (SVM).
21 November 2025 in Communications in Nonlinear Science and Numerical Simulation