LAUSR

Considering the variation in spectral response of plants due to unhealthiness effects, this study utilizes multispectral imaging to detect Greening disease of citrus trees. Low altitude multispectral images acquired in five discrete bands of R, G, B, Red Edge, and NIR by an imaging camera embedded on an unmanned aerial vehicle. Image features including 16 vegetation indices and 5 bands from spectral images are extracted. Support vector machine (SVM) used to classify the images using generated features in two steps. First, for determining trees from non-trees objects and then based on the output, healthy and diseased trees are classified. The obtained overall classification results based on check samples are 81.75% for SVM model which demonstrates that low altitude multispectral imagery has the potential to be applied for fast detection of Greening infected trees in citrus orchards.