Low rank and collaborative representation for hyperspectral anomaly detection via robust dictionary construction

Su, Hongjun; Wu, Zhaoyue; Zhu, A‐Xing; Du, Qian

doi:10.1016/j.isprsjprs.2020.09.008

Cited by 68 publications

(21 citation statements)

References 68 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…To reduce the computational complexity of the CRD, Ma et al [22] proposed a collaborative representation hyperspectral anomaly detection algorithm, which was a fast recursive algorithm, wherein two elementary transformation matrices were constructed based on the position of pixels, and furthermore, a recursive updating method was derived by using matrix inversion lemma, which increases the detection speed of the CRD to a certain extent. Su et al [23] further proposed an anomaly detection method by combining low rank representation (LRR) and CR theories, and meanwhile, introduced the global low rank and local collaborative properties to constrain the representation coefficient matrix.…”

Section: A Development Of Hyperspectral Anomaly Detection Methodsmentioning

confidence: 99%

See 1 more Smart Citation

Reweighted Nuclear Norm and Total Variation Regularization With Sparse Dictionary Construction for Hyperspectral Anomaly Detection

Wang

2022

IEEE J. Sel. Top. Appl. Earth Observations Remote Sensing

View full text Add to dashboard Cite

Hyperspectral anomaly detection is an important technique in the field of remote sensing image processing. Over the last few years, low rank and sparse matrix decomposition (LRSMD) has played an increasingly significant role in hyperspectral anomaly detection. The detection performance of the LRSMD-based anomaly detectors is primarily determined by prior constraints and the background dictionary construction method. To increase the detection accuracy, we proposed the reWeighted Nuclear Norm and total variation regularization with Sparse Dictionary construction for hyperspectral Anomaly Detection (WNNSDAD), which incorporated reweighted nuclear norm and total variation regularizations as the prior constraints into the LRSMD model, and constructed a sparse background dictionary without the need of clustering. Compared to the standard nuclear norm, the reweighted nuclear norm helped to overcome the challenge of an unbalanced penalty for a singular value and ensure a more effective low rank approximation. Simultaneously, total variation regularization was introduced as a piecewise smoothing constraint, which helped to maintain the spatial correlation of the hyperspectral image. Additionally, we proposed a background dictionary construction method, by which a relatively complete background dictionary could be obtained without clustering, and the background part could be represented more reliably. The experiments on seven real-world hyperspectral datasets show that in comparison to eight state-of-the-art anomaly detection methods, the proposed WNNSDAD method demonstrated greater accuracy.  Index Terms-Hyperspectral anomaly detection, low rank and sparse matrix decomposition (LRSMD), reweighted nuclear norm, total variation (TV), sparse background dictionary. I. INTRODUCTIONA hyperspectral image (HSI) comprises three-dimensional cube data. Two of these dimensions represent spatial position; and the other, the reflectance of ground objects in different bands. In comparison to multi-spectral images, HSIs have hundreds of continuous spectra, which could provide abundant Manuscript

show abstract

Section: A Development Of Hyperspectral Anomaly Detection Methodsmentioning

confidence: 99%

“…As there is no requirement for prior spectral information on anomalous target pixels, anomaly detection is more efficient in applications. Therefore, hyperspectral anomaly detection has been widely used in environmental monitoring [12], search and rescue [13], military reconnaissance [14], and other fields.…”

mentioning

confidence: 99%

Reweighted Nuclear Norm and Total Variation Regularization With Sparse Dictionary Construction for Hyperspectral Anomaly Detection

Wang

2022

IEEE J. Sel. Top. Appl. Earth Observations Remote Sensing

View full text Add to dashboard Cite

show abstract

“…Global low-rank features and local collaboration among atoms can be utilized by using low-rank representation (LRR) regularized by a collaborative term [20]. The resultant method is called collaborative and low-rank graph-embedding (CLGGE) for DR. Its objective function can be formulated as:…”

Section: A Clggementioning

confidence: 99%

“…Collaborative representation can be incorporated with global low-rank nature for the DR purpose. In [20], low-rank and collaborative representation is used for hyperspectral anomaly detection.…”

Section: Introductionmentioning

confidence: 99%

Collaborative and Low-Rank Graph for Discriminant Analysis of Hyperspectral Imagery

Shah

2021

IEEE J. Sel. Top. Appl. Earth Observations Remote Sensing

Self Cite

View full text Add to dashboard Cite

Sparse representation can be used for representation of a high-dimensional data into a low-dimensional subspace. Recently, sparse graph-based discriminant analysis that uses l1norm optimization has drawn much attention in dimensionality reduction of hyperspectral imagery. By combining low-rankness and sparsity, sparse and low-rank representation-based discriminant analysis (SLGDA) can effectively capture global and local data structures simultaneously. In this paper, collaborative and low-rank representation-based discriminant analysis (CLGDA) is proposed, which is different from the concept of sparse representation. More informative graph can be obtained in CLGDA with the combination of both collaborative representation (CR) and low-rank representation (LRR) because global data structure can be preserved by LRR and collaboration among within-class atoms is important than competition in sparse representation. Moreover, collaborative representation with l2norm regularization is computationally efficient and competitive to sparse representation. The experimental results conducted on three hyperspectral datasets demonstrate that the proposed CLGDA and its variants can provide better classification performance in comparison to SLGDA counterparts with lower computational complexity.

show abstract

“…yperspectral image provides abundant spectral information in hundreds of contiguous spectral bands [1]. This property has led to hyperspectral images widely applied in many applications [2], such as environmental monitoring [3], agriculture [4], mineral exploration [5], military [6].…”

Section: Introductionmentioning

confidence: 99%

Dynamic Ensemble Learning With Multi-View Kernel Collaborative Subspace Clustering for Hyperspectral Image Classification

Lü

et al. 2022

IEEE J. Sel. Top. Appl. Earth Observations Remote Sensing

Self Cite

View full text Add to dashboard Cite

Recently, a series of collaborative representation (CR) methods have attracted much attention for hyperspectral images classification. In this paper, two CR-based dynamic ensemble selection (DES) methods using multi-view kernel collaborative subspace clustering (MVKCSC) and random subspace multi-view kernel collaborative subspace clustering (RSMVKCSC) are proposed. In order to combine spectral and spatial information to construct a region of competence (RoC), the multi-view learning strategy is used in the general DES method. Compared with traditional clustering methods, the MVC can more effectively utilize multi-feature information. Moreover, a new method of constructing the Laplacian matrix using kernel collaborative representation coefficients is proposed for clustering based on subspace clustering and CR theory. This method is called MVKCSC, which can obtain the clustering results by using kernel CR self-representation coefficients. In addition, to increase diversity of samples, the random subspace method (RSM) and MVKCSC are combined for RMVKCSC. Moreover, the algorithm can obtain better clustering results by constraining samples and features simultaneously. The effectiveness of the proposed methods is validated using three hyperspectral data sets with few samples. The experimental results show that both DES-MVKCSC and DES-RSMVKCSC outperform their single classifier counterparts. In particular, the proposed methods provide superior performance compared with the state-of-the-art DES methods.

show abstract

Low rank and collaborative representation for hyperspectral anomaly detection via robust dictionary construction

Cited by 68 publications

References 68 publications

Reweighted Nuclear Norm and Total Variation Regularization With Sparse Dictionary Construction for Hyperspectral Anomaly Detection

Reweighted Nuclear Norm and Total Variation Regularization With Sparse Dictionary Construction for Hyperspectral Anomaly Detection

Collaborative and Low-Rank Graph for Discriminant Analysis of Hyperspectral Imagery

Dynamic Ensemble Learning With Multi-View Kernel Collaborative Subspace Clustering for Hyperspectral Image Classification

Contact Info

Product

Resources

About