Hyperspectral Change Detection Using Mahalanobis Distance
Master of Science in Electrical Engineering (MSEE)
Electrical and Computer Engineering
Abstract Through the past couple of decades, Hyperspectral Change Detection (HCD) has proven to be an encouraging approach for autonomously detecting subtle targets amongst complex backgrounds. However, the detection of the targets can become complicated by natural changes and variations occurring in the background clutter, causing a false report of a target change, commonly known as a “false alarm”, or the act of “missing” a target change. These types of variations can range from changes in illumination, such as shadowing and time of day, to seasonal alterations, such as the difference of coloring amongst vegetation. This thesis focuses on the technique of increasing the success rate of detecting man-made changes in hyperspectral images (HSIs) while decreasing the false-alarm rate without raising the target-miss rate, which has proven to be a difficult task. Due to their computational simplicity and mathematical tractability, linear detectors are the main focus of this study. Through the use of linear detectors and the Central Limit Theorem, it is assumed that the affine transforms that are applied to the images produce approximately Gaussian HSIs, even though we briefly consider non-Gaussian situations. Three separate algorithms are analyzed and compared, based on linear estimation theory: Canonical Correlation Analysis (CCA), Chronochrome (CC), and Covariance Equalization (CE). Mathematical expressions of the Mahalanobis distance are derived for each of these algorithms. Further analysis of change detection reveals the connection between the Mahalanobis distance and the optimum variance estimator for the assumed standard Gauss model, which is shown to hold approximately for linear detectors. Finally, the proposal for a new scheme in change detection is shown to be effective via both mathematical analysis and simulation results.
Document Availability at the Time of Submission
Secure the entire work for patent and/or proprietary purposes for a period of one year. Student has submitted appropriate documentation which states: During this period the copyright owner also agrees not to exercise her/his ownership rights, including public use in works, without prior authorization from LSU. At the end of the one year period, either we or LSU may request an automatic extension for one additional year. At the end of the one year secure period (or its extension, if such is requested), the work will be released for access worldwide.
Baines, Marcus DeJuan, "Hyperspectral Change Detection Using Mahalanobis Distance" (2011). LSU Master's Theses. 936.