An Improved Hyperspectral Image Anomaly Detection Algorithm using Low-Rank Representation

Abstract

Anomaly detection in hyperspectral images has drawn much attention in recent years. In order to provide a high-quality background dictionary for low-rank representation-based anomaly detector, from the perspective of dictionary learning, an anomaly detection method based on low-rank representation with an online-learned double sparse dictionary is proposed. Firstly, the double sparsity structure is adopted to the dictionary learning model to enhance the adaptivity. Next, to improve the dictionary training efficiency, the double sparse dictionary structure is modified and a corresponding online dictionary learning algorithm is proposed. The experimental results on five real-world hyperspectral datasets show that our method can achieve a reliable anomaly detection result and the background suppression performance is satisfying.