Hyperspectral Anomaly Detection via Sparse Dictionary Learning Method of Capped Norm

Abstract

Hyperspectral anomaly detection is a research hot spot in the field of remote sensing. It can distinguish abnormal targets from the scene just by utilizing the spectral differences and requiring no prior information. A series of anomaly detectors based on Reed-Xiaoli methods are very important and typical algorithms in this research area, which generally have the hypothesis about background subject to the Gaussian distribution. However, this assumption is inaccurate to describe a hyperspectral image with a complex scene in practice. Besides, due to the unavoidable existence of abnormal targets, background statistics will be affected which will reduce the detection performance. To address these problems, we propose a sparse dictionary learning method by using a capped norm to realize hyperspectral anomaly detection. Moreover, a new training data selection strategy based on clustering technique is also proposed to learn a more representative background dictionary. The main contributions are concluded in threefold: 1) neither making any assumptions on the background distribution nor computing the covariance matrix, the proposed method is more adaptive to all kinds of complex hyperspectral images in practice; 2) owing to the good qualities of the capped norm, the learned sparse background dictionary is resistant to the effect of anomalies and has stronger distinctiveness to anomalies from background; 3) without using the traditional sliding hollow window technique, the proposed method is more effective to detect different sizes of abnormal targets. The extensive experiments on four commonly used real-world hyperspectral images demonstrate the effectiveness of the proposed method and show its superiority over the benchmark methods.

From Model-Based Optimization Algorithms to Deep Learning Models for Clustering Hyperspectral Images

“…Motivated by the success of dictionary learning in signal processing and highdimensional data analysis [144][145][146][147][148][149], recent works [76][77][78][79] replace the self-representation dictionary with an adaptive dictionary that is leaned from the input data, resulting in computationally efficient clustering models. The developed models often consist of three steps, joint dictionary learning and sparse coding, similarity matrix construction and spectral clustering.…”

Subspace Clustering for Hyperspectral Images via Dictionary Learning With Adaptive Regularization

“…It is thus of interest to construct efficiently compact dictionaries to model the underlying low-dimensional subspaces of a HSI. It has been demonstrated that learning a dictionary from data instead of using a predefined one can effectively improve the performance of data analysis [46][47][48][49][50][51][52]. This motivates us to learn a compact dictionary to model the low-dimensional subspaces of HSIs in our subspace clustering method.…”

Multiscale Graph Sample and Aggregate Network With Context-Aware Learning for Hyperspectral Image Classification

“…(2) shows multi-scale graph learning, which is the input of the subsequent networks. Then a SAGE layer is adopted to reconstruct the multi-scale graph, the reconstructed graph can be expressed as follows (6) Where denotes SAGE mechanism (Algorithm 1), means graph reconstruction, is the weight matrix of the network, which can be optimized based on backpropagation loss as network training. In the training process, the relationship between nodes in graphSAGE layer generated graph will also change as the classification target changing, so that we can construct different branches graph node relationships according to different classification targets.…”