Maximum relevance and class separability for hyperspectral feature selection and classification

Regarding a growing interest into exploiting hyperspectral images in the plethora of applications such as chemical material identification, agricultural crop mapping, military target detection and etc., myriad approaches have been introducing to interpret and analyze such data. In this paper, I am going to propose a novel method using the combination of two conventional method. Firstly, I use an evolutionary algorithm i.e., multi-objective particle swarm optimization (MOPSO) to select a predefined number of features (spectral bands) and then a well-known classifier i.e., support vector machines (SVMs) is deployed for classification.