Pattern Recognition based on Fuzzy Inference of Spectral-Spatial Features

Soft computing based on spectral spatial features are introduced by this pater to analyze hyperspectral images. In the proposed pattern recognition algorithm, after pre-processing, three pre-processing steps are applied, including the construction of initial probability maps, filtering of probabilistic maps, and maximizing the probability to the soft computing classification class. In this article, some of the spectral and spatial features of pre-processed remote sensing images are offered to the Mamdani fuzzy inference system. The optimal features for applying to the fuzzy inference system are selected using the genetic algorithm and SVM. The efficiency of the proposed pattern recognition algorithm is assessed using some real datasets and several evaluation criteria. The empirical results efficiently show the performance of the proposed method in classifying hyperspectral and computational intelligence algorithms.