Classificação de tipos de bananas utilizando imagens hiperespectrais e redes neurais perceptron multicamada

Abstract

The purpose of the present work is to apply spectroscopic techniques together with pre-processing methods to organize data in the classification of samples, specifically, banana samples. The objective is to highlight the importance and effectiveness of hyperspectral images with pre-processing using K-means to group pixels with similar characteristics, differentiating samples from the background. Subsequently, the multilayer perceptron algorithm is applied to classify banana species, such as apple banana, dwarf banana and silver banana, ensuring a classification accuracy of 75%. This demonstrates the algorithm's ability to improve its learning. It is concluded that the set of techniques used in this work can be used to classify different types of samples in any area.