Target detection in hyperspectral imagery based on linear mixing model reconstructed from measurements

There presents a detection algorithm based on spectral mixing model reconstructed from measurement in this paper in order to detect unknown targets in unknown environment. Firstly, we project the hyperapectral imagery to suppress the background interference in order to search target spectral more accurately. Then, we estimate the spectral subspace and construct a spectral mixing model reconstructed from measurements. And based on the proposed spectral mixing modeling, we project the hyperspectral imagery, which suppress spectral signatures of background and improve the SNR, in order to increase the detection power. Finally, the Signal to Local Clutter RMSE (SLCR) and Peak Signal to Local Clutter Mean Ratio (PSLCMR), which is proposed, are used to evaluate the detection. Theoretic analysis and the results of experiment on visible/near-infrared hyperspectral imagery verify the effectiveness of the algorithm.