Research on Hyperspectral Anomaly Target Detection Algorithm Based on Dual Window Subspace Projection

Abstract

As an important application field of hyperspectral remote sensing technology, hyperspectral abnormal target detection is widely used in the fields of environmental monitoring, aerospace and military remote sensing. Based on orthogonal subspace projection (OSP) and local orthogonal subspace projection (LOSP) anomaly target detection algorithms, this paper proposes a hyperspectral anomaly target detection algorithm (DLOSP) that constructs a dual-window subspace anomaly detection operator using the neighborhood of detecting pixel. The experimental results show that the algorithm proposed in this paper has higher detection accuracy, lower virtual scene rate, low computational complexity, and detection time is much lower than RXD algorithm, which is suitable for real-time hyperspectral anomaly target detection system.