A New Processing Method to Segment Olive Trees and Detect Xylella Fastidiosa in UAVs Multispectral Images

Abstract

In this paper, a new approach for fast detection of Xylella fastidiosa bacterium symptoms on olive trees is presented. Images are taken using a multirotor unmanned aerial vehicle (UAV) equipped with a multispectral camera. A new segmentation algorithm to recognize trees is applied and images are then classified using linear discriminant analysis. It has been applied to selected sites in the Southern Italy where multispectral images of olive orchards have been acquired. The developed algorithm seems to be very promising thanks to its high mean Sørensen-Dice similarity coefficient, which demonstrates the feasibility of a correct tree individuation, and its sensitivity in detecting infected trees.