Extração de Características para Geração de um Classificador para Detecção Precoce do HLB em Citros

Diego Carlos Pereira da SilvaPatrícia Pedroso Estevam RibeiroLúcio André de Castro JorgeMaria Stela Veludo de PaivaDébora Marcondes MiloriDanilo Scavacini GonçalvesCamila Miranda CarvalhoAndré Leonardo VenânicoFabíola Manhas Verbi Pereira

Greening or Huanglongbing (HLB) is one of the most serious threats for citrus production worldwide. São Paulo, Brazil is the most important citrus producers and, is making efforts for citrus diseases control. The bacterial pathogen is, mainly, under suppression control, applied by eradication of symptomatic and no-symptomatic plants. In this way, the detection and diagnostic of the related symptoms and, consequently, the eradication of the citrus trees are essential for higher economical losses prevention. In this way, our goal is to develop a new optical and data mining technique, applied in field conditions, to detect citrus diseases using a fluorescence imaging system. It was used a system that perform image of the chlorophyll fluorescence in the whole leaf. It was collected images of 3 leaves from 120 different greenhouse rootstocks. It has obtained fluorescence images and spectral data of citrus healthy leaves and contaminated leaves with HLB. In this paper it was described a new method to extract attributes by data fusion, carrying out different analysis to try to discriminate between safe and diseased leafs. The results show that the method is very accurate for HLB ultimately detection.

Caso o link acima esteja inválido, faça uma busca pelo texto completo na Web: Buscar na Web

Biblioteca Digital Brasileira de Computação - Contato:
     Mantida por: