Target Detection Algorithms in Hyperspectral Imaging Based on Discriminant Analysis

Target detection algorithm in hyperspectral imaging detects a certain material in a hyperspectral image using a known spectral signature of the material. Conventional algorithms for target detection assume that there is only one known target spectrum so target statistics cannot be estimated. Discriminant analysis is designed for classification, but this paper analyzes the performance of discriminant functions for target detection. The discriminant functions have been modified for target detection and uses simulated target spectra with different amount of random noise. Experimental results show that the algorithms can work well within a certain amount of noise.