Adaptive Feature Selection for Hyperspectral Data Analysis Using a Binary Hierarchical Classifier and Tabu Search

High dimensional inputs coupled with scarcity of labeled data are among the greatest challenges for classification of hyperspectral data. These problems are exacerbated if the number of classes is large. High dimensional output classes can often be handled effectively by decomposition into multiple two(meta)class problems, where each sub-problem is solved using a suitable binary classifier, and outputs of this collection of classifiers are combined in a suitable manner to obtain the answer to the original multi-class problem. This approach is taken by the binary hierarchical classifier (BHC). The advantages of the BHC for output decomposition can be further exploited for hyperspectral data analysis by integrating a feature selection methodology with the classifier. Building upon the previously developed best bases BHC algorithm with greedy feature selection, a new method is developed that selects a subset of band groups within metaclasses using reactive tabu search. Experimental results obtained from analysis of Hyperion data acquired over the Okavango Delta in Botswana are superior to those of the greedy feature selection approach and more robust than either the original BHC or the BHC with greedy feature selection. INTRODUCTION For classification of hyperspectral data, there are potentially hundreds of correlated inputs which may result in unstable estimates of parameters, particularly when there is a small quantity of training data. This problem is commonly addressed using some form of feature extraction, such as feature subset selection. The computational complexity of optimal feature selection methods has forced acceptance of heuristic techniques that find good, near-optimal subsets in relatively short computational times. A comparative study of several of the well-known optimal and sub-optimal feature selection algorithms [e.g. variations of Sequential Forward/Backward Selection (SFS/SBS), Branch and Bound and relaxed Branch and Bound] is contained in [1]. Other ACKNOWLEDGMENT This work was supported by the National Aeronautics and Space Administration under the EO-1 program (NCC5-463). approaches include genetic algorithms [2], simulated annealing [3], and the Tabu Search (TS) metaheuristic [4]. In this study, a new model is developed which incorporates the use of TS with the multiclassifier system known as the Best Bases Binary Hierarchical Classifier (BB BHC) [5-8] for analysis of hyperspectral data. The primary goal in development of the BHC was output decomposition for problems with a medium to large number of classes. While the classification accuracies obtained from the BHC are typically good, problems are encountered if the input is highdimensional, and the amount of training data is limited.