A Constrained Sparse-Representation-Based Binary Hypothesis Model for Target Detection in Hyperspectral Imagery

Abstract

In this paper, we propose a novel constrained sparserepresentation-based binary hypothesis model for target detection in hyperspectral imagery. This model is based on the concept that a target pixel can only be linearly represented by the union dictionary combined by the background dictionary and target dictionary, while a background pixel can be linearly represented by both the background dictionary and the union dictionary. To be physically meaningful, the non-negativity constraint is imposed to the weight vector. To suppress the target signals in the background dictionary, the upper bound constraint is also imposed to the weight vector. These upper bounds are adaptively estimated by the similarities between the atoms in the background dictionary and target. Then, the weight vectors under different hypotheses are recovered by a fast coordinate descent method. Finally, both the residual difference and weight difference between the two hypotheses are used to perform the target detection. An important advantage of the proposed method is the robustness to varying target contamination. Extensive experiments conducted on real and synthetic hyperspectral datasets have demonstrated the superiority of the proposed detector in detection performance and computational cost. Specifically, for the Avon dataset, our method achieves the highest area under the receiver operating characteristic (ROC) curve of 99.91%, and achieves the shortest runtime of 109.76 s.

TANet: Transformer‐based asymmetric network for RGB‐D salient object detection

“…We also evaluate traditional HTD methods [10]- [12], [21], [23]- [25], [29], [30], [35], [36], [56]. For each type of object, we provide 20 prior spectral curves.…”

Hyperspectral Detection and Unmixing of Subpixel Target Using Iterative Constrained Sparse Representation

“…The continuous spectral signature for different materials is of great help in target detection and component identification. The discriminative property leads to various applications in HSI remote sensing: denoising [2], spectral superresolution [3], target detection [4], [5], spectral unmixing [6]- [8], and dimensionality reduction [9]. Among these applications, target detection and spectral Manuscript received XXX, 2021; revised XXX.…”

“…Recently, sparse representation (SR) based techniques have been successfully applied to HSI target detection [5], [17]- [19]. The basic SR-based detector [17] represents each test pixel as a sparse linear combination of atoms from the background dictionary or target dictionary, and then detects the targets by using representation residuals.…”

“…The hybrid sparsity and statistics (HSS) detector [19] combines sparse representation with statistical distribution theory, and also applies a purification process to handle target contamination in the background dictionary. The constrained sparse representation-based binary hypothesis detector [5] imposes an upper bound constraint to the abundance vector. Then these upper bounds are automatically estimated by the background information and prior target information.…”

The Detection of Concealed Targets in Woodland Areas Using Hyperspectral Imagery

“…With tens to hundreds contiguous spectral bands, hyperspectral imaging technology is now able to differentiate subtle spectral differences between land-cover objects, thus allowing for advanced material identification and separation techniques such as target detection (Q. Ling, 2019). Accordingly, hyperspectral technology has been successfully implemented for several applications areas, including, e.g., medicine, environment, public safety, and defense (Stefanou, Kerekes, 2010).…”