The utility of hyperspectral data to detect and discriminate actual and decoy target vehicles

Abstract

The objective of this work is to evaluate the utility of hyperspectral signature data in satisfying time-sensitive intelligence requirements. This work is conducted in support of the Hyperspectral MASINT support to Military Operations (HYMSMO) program. Data are used from the Hyperspectral Digital Imaging Collection Experiment (HYDICE) imaging spectrometer using the 0.4 um to 2.5 um wavelength range. Operation Forest Radiance I was the third in a series of HYMSMO- sponsored collection and exploitation experiments, and the data set analyzed herein was derived from this effort. The first phase of the Forest Radiance experiment emphasized the collection of spectra from a suite of overtly exposed mobile vehicles, decoys, and target panels. Analysis shown here was conducted to determine if it is possible to detect and discriminate real and decoy vehicles. The Low Probability of Detection (LPD) and Spectral Angle Mapper (SAM) anomaly detection and classification algorithms are applied to the data set being analyzed. The LPD algorithm performs well at detecting residual spectra, but produces a significant number of false alarms. The SAM technique is equally successful at detecting residual spectra and proves to have an advantage over the LPD when it comes to obviating misidentifications. This thesis shows that detection and discrimination of mobile vehicles (HMMWVs) and decoys in a natural grass environment is possible using this technology