DETECTING CITRUS HUANGLONGBING IN BRAZILIAN ORCHARDS USING HYPERSPECTRAL AERIAL IMAGES
Keywords: Citrus Greening, Remote Sensing, Precision Agriculture, Health Map
Abstract. Brazil is one of the world leaders in citrus plantation, and the production of orange juice is economically important in the export scenario, being regarded as a fundamental agricultural commodity in Brazil. The worst citrus disease is Greening, or Huanglongbing (HLB), a bacterial disease which cannot be cured and to which no plant variety is immune or resistant. Currently, control of HLB is through the inspection of orchards and the immediate elimination of plants displaying HLB symptoms, plus chemical or biological control of the insect vector (psyllid). The HLB disease has high economic impact on Brazilian and world citriculture, due to the extreme damage to crops. Based on remote sensing techniques, the mapping of diseased and healthy citrus plants from hyperspectral images was carried out, generating a product that could help in the monitoring and control of HLB in Brazilian citrus orchards. The methodology to produce a health map goes through the following stages: aerial image acquisition, radiometric field measurements, hyperspectral cube orientation, and data analysis for detection of HLB in citrus. The field survey was performed in Guacho Farm- Brazil and hyperspectral images were acquired by a Rikola camera onboard a light aircraft, obtaining images with 0.50 m of Ground Sample Distance (GSD). The hyperspectral cubes were classified with Spectral Angle Mapper (SAM) algorithm to produce the health map. Plants infected with HLB were detected with an accuracy 61.2%, the validation of the health map was verified by samples were analysed in the laboratory to confirm HLB.