Anomaly detection using subspace band section based RX algorithm

Large amount of spectral wavebands of hyperspectral remote sensing data provide an abundant and valuable information for classification and detection, but at the same time, this will lead to redundancy and complexity of the data. In this paper, we proposed a subspace band section based RX algorithm for hyperspectral imagery (HSI) anomaly detection. Firstly, the adaptive subspace decomposition (ASD) algorithm divides the HSI data of high dimension to subsets of low dimension according to the correlation between spectral wavebands. Then principle component analysis (PCA) is used in each subspace to select certain principle components for RX detection. At last, fuse the results of each subspace and get the anomaly elements through a certain threshold. The experimental results show that the proposed algorithm can achieve high detection performance.