One dimensional fringe-adjusted joint transform correlation and adaptive band selection in hyperspectral imagery

In this thesis, a novel one-dimensional fringe-adjusted joint transform (1D-FJTC) correlation technique and an adaptive band selection technique is proposed for detecting small objects involving only a few pixels in hyperspectral imagery. In the 1D-FJTC, one dimensional (1D) spectral signatures from unknown hyperspectral imagery are correlated with the known reference signature. This technique can detect both single and/or multiple objects in constant time while accommodating the in-plane and out-of-plane distortion. In the adaptive band selection technique, the selection of the lowest number of spectral bands is ensured using Mahalanobis distance where noise is added to the maximum likelihood estimate of the covariance matrix. The addition of extra noise ensures that the band selection is robust. Due to the spectral variability, band selection based only on the maximum likelihood estimates is not adequate and the bands which can withstand extra noise should be chosen. The combination of the aforementioned techniques is also proposed to reduce computation complexity and dimensionality of the datacube. Test results using real life hyperspectral image datacubes are presented to verify the effectiveness of the proposed techniques.