Extraction of Peronospora infection through the analysis of hyperspectral data

In the last years, a great interest in precision agriculture field has developed. Technology strongly supports this strategy employing innovative concepts and tools to assist farmers in agricultural processes management. Sensors, typically mounted on UAVs (Unmanned Aerial Vehicles) or UGVs (Unmanned Ground Vehicles), play a fundamental role in the described scenario since they allow an extensive evaluation of the framed objects with a high detail level. In this context, the present thesis aims to analyze the data gathered through the hyperspectral camera Rikola in an Astigian vineyard. The acquired images show vine leaves affected by a typical vine disease, Peronospora, caused by a fungus (Plasmopara viticola). The effect is the generation of spots on the leaves with the consequent leaves necrosis, but damages can extend to branches and grapes. This represents a loss of resources and therefore a reduction of production. For these reasons, it is important to prevent or at least limit the expansion of Peronospora in vineyards. Thanks to the continuous technological development, plants health control and autonomous identification of several phenomena are becoming key operations. This thesis work is propaedeutic to the explained goals: an initial calibration phase is applied in order to remove the influence of external factors on acquired data. Then calibrated images are elaborated with machine learning algorithms to perform a classification process, which leads to obtain a classifier able to identify Peronospora effects on vine leaves and that can be employed in different applications.