Target detection and classification at Kernel Blitz 1997 using spectral imagery

Abstract

Data collected from the Hyperspectral Digital Imagery Collection Experiment (HYDICE) were analyzed in this thesis to determine the feasibility of wide area detection and classification of target materials in the visible to short wave infrared region. This study used detection algorithms such as spectral angle mapper and matched filter for target detection. Parallelepiped and Maximum Likelihood routines were used to classify the image data for subsequent analyses and comparisons. Effects on data due to altitude variation of the sensor were analyzed using histograms, differencing, and principal component transforms. Data images of the Camp Pendleton airfield used for comparison analyses were obtained from two different altitudes, 5,000 feet and 10,000 feet. Results showed target detection and classification had no strong dependence on altitude.