Advanced Signature-Matched Hyperspectral Change Detection

Gitam Technologies Inc. (GTI) in collaboration with Professor John Kerekes of Rochester Institute of Technology, propose to kernelize the Covariance Equalized Change Detection algorithm for Hyperspectral imagery. The primary focus will be to develop signature-based change detection with the capability to locate particular signatures in an observation scene, where prior knowledge about the target or the background is available from previous pass. Over the past decade, Kernel-based learning concepts have established as a powerful optimization tool when gaussian and linearity assumptions are invalid. In kernel-based methods, learning is implicitly performed in a high-dimensional feature space where high order correlation or nonlinearity within the data is exploited. The kernel concept has seen vast applications in wide variety of detection and recognition problems, including traditional detection and classification, change detection and anomaly change detection. However, although Covariance Equalized Change Detection (CECD) has seen several useful applications, the CECD concept has not been formalized within the Kernel framework. Hence, the main goal of this project is to study the effectiveness of the entire class of Kernels for CECD for HSI data. The Kernelization approach will be initially applied to the chronochrome algorithm and the covariance equalization algorithm both with and without matched filtering.