Anomaly Detection in Hyperspectral Data with Matrix Decomposition

The role of anomaly detection in hyperspectral imaging is increasingly important. Traditional anomaly detection methods mainly extract information from background images. They use this information to find the difference between anomalies and background. Using generally background information for detecting anomalies and modeling background can cause background contamination with anomaly pixels. However, Low Rank and Sparse Matrix Decomposition (LRaSMD) based methods can solve this problem due to using both background and anomaly information. In this study, an LRaSMD based anomaly detection method is adopted. According to the experimental results, the proposed method shows better performance than other state-of-art methods.