Low-rank & sparse matrix decomposition and support vector machine for hyperspectral anomaly detection

Abstract

Due to the limited resolution of hyperspectral sensors, anomalous targets expressed with subpixels are often mixed with nonhomogeneous backgrounds. This fact makes anomalies difficult to be distinguished from the surrounding background. From this perspective, we propose a novel hyperspectral anomaly detection (AD) algorithm based on low-rank & sparse matrix decomposition (LRaSMD) and support vector machine (SVM). First, based on the LRaSMD technique, the Go decomposition (GoDec) model is utilized to decompose the original image into three components: background, anomalies and noise. In this way, the robust background can be obtained. Subsequently, a clustering algorithm is employed to pick some obvious anomalies. Accordingly, we use both samples of background and anomaly to train an SVM model. The original dataset is sent into the SVM model and both anomalous components and background components can be classified. Experiments on a synthetic hyperspectral image validate the performance of the proposed method.