Anomaly Detection Improvement Using Sparse Representation and Morphological Profile

A sparse based anomaly detection method using spatial information is proposed for hyperspectral images in this paper. Sparse representation provides an appropriate feature space for discrimination between anomalous targets and background. In addition, differential morphological features can result in valuable spatial information about anomalies. So, the simultaneous use of sparse representation and differential morphological profile (DMP) increases anomaly detection rate and decreases the false alarm rate. The experiments shows about 30% incensement in detection accuracy compared to the background joint sparse representation (BJSR) method.