Multiobjective PSO for hyperspectral image clustering

In this chapter, a multiobjective particle-swarm optimization approach is presented as an answer to the problem of hyperspectral remote sensing image clustering. It aims at simultaneously solving the following three different issues: (1) clustering the hyperspectral cube under analysis; (2) detecting the most discriminative bands of the hypercube; (3) avoiding the user to set a priori the number of data classes. The search process is guided by three different statistical criteria, which are the log-likelihood function, the Bhattacharyya distance, and the minimum description length. Experimental results clearly underline the effectiveness of particle-swarm optimizers for a completely automatic and unsupervised analysis of hyperspectral remote sensing images.