Constrained-Target Band Selection With Subspace Partition for Hyperspectral Target Detection

Abstract

Hyperspectral target detection is widely used in both military and civilian fields. In practical applications, how to select a low-correlation and representative band subset to reduce redundancy is worth discussing. However, most of the existing band selection (BS) methods usually select bands according to the statistics or correlation, which neglect the spectral characteristics of desired target and are not specially designed for target detection. Therefore, this article proposed a novel BS method, called constrained-target BS with subspace partition (CTSPBS), to select an optimal subset with low internal correlation and strong target representability for target detection task. By using a specially designed subspace partition method based on correlation distance (CDSP), CTSPBS divides the hyperspectral bands into several unrelated subspaces. Then, according to certain constrainedtarget band prioritization (BP) criterion, the band with highest priority in each subset is selected to form the optimal subset for specific target. Correspondingly, two versions of proposed method, minimum variance BS with CDSP (CDSP_MinV) and minimum variance BS with CDSP (CDSP_MaxV) are derived to implement CTSPBS. Extensive experiments on three public hyperspectral datasets demonstrate that the proposed method exhibit more robust and effective performance than several state-of-the-art methods. Finally, this article focuses on the difficulty of marine benthos detection in mariculture application, and proves the feasibility of the proposed method.

MOBS-TD: Multiobjective Band Selection With Ideal Solution Optimization Strategy for Hyperspectral Target Detection

“…Instead of following the idea of CBS which constrains a band of interest, Wang et al [40] proposed the constrained-target band selection (CTBS) by constraining a target of interest, which makes full use of the known target features. Afterwards, Sun et al [41] proposed the constrained-target band selection with subspace partition (CTSPBS), which further improves the detection performance by reducing redundancy. Shang et al [8] proposed the target-constrained interference-minimized band selection (TCIMBS), focusing on how to eliminate the effect of uninteresting targets with similar spectra on detection of interesting targets.…”

“…VD is defined as the minimum number of spectrally distinct signal sources that characterize the hyperspectral data from the perspective view of target detection and classification. Many recently proposed BS methods for target detection, such as CTBS [40], CTSPBS [41], TCIMBS [8], and TOMOBS [45], use VD to determine the number of bands and validate its effectiveness. Thus, we adopt the noise whitened Harsanyi-Farrand-Chang (NWHFC) [58] method to estimate VD and then determine the BS n .…”

Anomaly‐background separation and particle swarm optimization based band selection for hyperspectral anomaly detection

“…Sun et al. [19] proposed a subspace partitioning method based on correlation distance, which selects the representative bands from multiple weakly correlated subsets to form an optimal band subset with low correlation and strong target expression capability. This method was successfully applied to a real underwater target detection scenario, realizing the transformation from theory to practical application.…”

“…In HSI processing, classification and target detection are two distinct research areas with different focuses due to the significant differences in the scales of desired ground objects. Classification-oriented BS methods tend to measure band quality in terms of global features or informativeness, while target detection-oriented BS methods are more concerned with local features or target-background separation [18][19][20]. Due to the difficulty in obtaining spectral information of target ground objects and the relative lack of available a priori information to support relevant studies to obtain reliable results, there are relatively few studies on target detection.…”

“…For supervised target detection, Xu et al [23] defined a novel objective function in the search process of particle swarm optimization (PSO) algorithm, that is, maximum-submaximum ratio, to measure the separability between target and background. Sun et al [19] proposed a subspace partitioning method based on correlation distance, which selects the representative bands from multiple weakly correlated subsets to form an optimal band subset with low correlation and strong target expression capability. This method was successfully applied to a real underwater target detection scenario, realizing the transformation from theory to practical application.…”