Image-Based Detection of Nutrient Deficiency in Vine Leaves

This thesis is the first step of a larger exploration study that aims to exploit Image Analysis in order to detect the presence of pathological conditions affecting vine plants. Vines can suffer different kinds of diseases, mainly resulting from deficiency of mineral nutrients or due to infestation by parasites. Nutrients play a fundamental role in grapevine development and their lack can be related to a peculiar colouration of the leaves, diverging from the typical green. The main colours associated to diseases are: yellow, being a symptom of chlorosis (insufficient chlorophyll) and red/purple, hinting at excessive content of carbohydrates and the activation of the anthocyanin biosynthesis. The part of the study this thesis focuses on, has the purpose to analyse the visual effects of mineral diseases and to design a feasible technique to enable vines classification as ill or not ill based on images of their leaves. The task has been addressed by means of a combination Pattern Recognition techniques including image processing an classification steps. The proposed pipeline has been evaluated over a sizeable database of vine leaves representing a significant range of different conditions.