Hyperspectral Anomaly Detection Based on Chessboard Topology

Without any prior information, hyperspectral anomaly detection is devoted to locating targets of interest within a specific scene by exploiting differences in spectral characteristics between various land covers. Traditional methods originated from the signal processing perspective, and most of them rely heavily on specific model assumptions. Because of the model-driven attributes, such methods cannot mine the deep-level features of data to adapt to the variability of scenes and cannot fully extract the information of land covers contained in images to accurately separate anomalies from the background. By independently designing a chessboard-shaped topological framework that avoids making any distribution assumptions but directly mines high-dimensional data features to break through the limitations of traditional detectors, this article proposes a novel chessboard topology-based anomaly detection (CTAD) method to dissect images and extract detailed information of land covers adaptively, thereby enabling highly accurate detection. Extensive experimental results on hyperspectral images (HSIs) in real scenes demonstrate that the proposed CTAD can be adapted to the variability of scenes by autonomously learning data features and exhibiting strong generalization and detection capabilities, facilitating practical applications.

I. Introduction

With the rapid development of hyperspectral remote sensing imaging technology, the quality of the obtained hyperspectral image (HSI) data has been significantly improved [1], [2]. The abundant and detailed information on land covers contained in images puts forward higher requirements for data mining and information extraction techniques [3], [4], [5]. As one of the most important research and application directions in hyperspectral remote sensing, target detection is devoted to locating objects of interest in an imaged scene with spectral characteristics of land covers [6], [7]. Based on available prior spectral information, target detection can be divided into two categories: supervised matching detection and unsupervised anomaly detection [1], [8]. In practical applications, it is very likely that there is a lack of informative and complete spectral libraries or accurate reflectance inversion algorithms [9], [10], coupled with the constraints of spectral measurement conditions [11], all of which make matching detection encounter serious limitations. While the operators adopted in anomaly detection algorithms do not require any prior spectral information of targets or background [8], [12], hence they have been widely used in such cases. As one of the research hotspots in HSI processing, unsupervised anomaly detection is of great significance in practical applications [13], [14].