Distribution-Level Hyperspectral Target Detection Under Mixture of Gaussian

View at publisher Summarise with VIRAT
Authors: Yubo Ma, Yinuo Zhao, Yiyun Gou, Qingke Zou, Siyu Cai
Cited by: 0
References: 18

Summary

No summary yet. VIRAT can read this paper and write a plain-language summary of its aim, methods, key findings and limitations.

Summarise with VIRAT Free to try with your Science Catalogs account.

Abstract

Statistical inference-based hyperspectral target detection, which provides a distribution-driven perspective for distinguishing the target from background, has recently attracted much attention. Most of the existing statistical detectors concentrate on modeling hyperspectral image (HSI) with an appropriate distribution or designing a better hypothesis test. However, the intrinsic relations of pixel–distribution and distribution–distribution have been ignored. Aiming to address this problem, a novel distribution-level target detection (DLTD) method for HSI is proposed, which incorporates distribution-level detection into pixel-level one and possesses statistical interpretability. Specifically, an HSI is statistically modeled as a mixture of Gaussian, where each Gaussian component characterizes an endmember. To explore detection at distribution level, the distance between indeterminate distribution and target one is measured by Kullback-Leibler divergence. Similarity based on this divergence is the detection value of the distribution. Then, the detection output is a weighted average of distribution-level detection values, where the weight is posterior probability. Experiments on four HSIs demonstrate superiority of the proposed DLTD method over comparative target detectors.

Fields of Study & Subjects

Remote-Sensing Image Classification Infrared Target Detection Methodologies Advanced Image Fusion Techniques

Publication details

Journal / source
IEEE Geoscience and Remote Sensing Letters
Volume
23
Pages
1–5
Publication date
1 January 2026
Language
EN