Amélioration de la surveillance épidémiologique du « greening » des agrumes infectés par le Huanglongbing à La Réunion par Spectroscopie Proche Infra-Rouge
Mémoire de stage de fin d’étude pour l’obtention du diplôme de Master 2, Ingénieur Agronome, Spécialisation GEEFT - Gestion Environnementale des Écosystèmes et Forêts Tropicales
2021-07-08
Abstract
Huanglongbing (HLB) is a bacterial disease causing dieback in citrus fruits. There has been a resurgence of this disease on Reunion Island since 2012, which threatens citrus orchards. The conventional detection techniques for disease surveillance by PCR are expensive, time consuming and cannot be carried out on a large scale. This is where the technology of spectral imaging analysis comes in. This technology is based on the analysis of the spectral signature that a medium emits in response to light exposure. This method is non-destructive, in addition to being relatively inexpensive. The analyzes made it possible to highlight a fairly weak influence of the plots and varieties on the wavelengths where HLB is detected. This makes the effect of HLB detectable without noise and therefore usable for diagnosis. Added to this is a processing of spectral imaging data by machine learning in order to predict the status of trees vis-à-vis the disease within citrus plots. This model is promising with a prediction quality of 92.6% for the Partial Least Squares (PLS) method on a training basis of 8400 reflectance spectra.