A Band Selection Method For Hyperspectral Image Classification Based On Cuckoo Search Algorithm With Correlation Based Initialization

Abstract

Band selection is an effective way to reduce the size of hyperspectral data and to overcome the curse of dimensionality problem in ground object classification. This paper presents a band selection method based on modified cuckoo search optimization with correlation-based initialization. The cuckoo search is one of the most effective and popular metaheuristic algorithms with efficient optimization capabilities for band selection. However, it can easily fall into local optimum solutions. In order to avoid falling into a local optimum, an initialization strategy based on correlation is adopted instead of random initialization to initiate the location of nests. Experimental results with AVIRIS Indian Pines data show that the proposed method obtains overall accuracy of 82.83% which is higher than the original binary cuckoo search algorithm, Genetic Algorithm (GA) and Particle Swarm Optimization (PSO).