The Automation of Hyperspectral Training Library Construction: A Case Study for Wheat and Potato Crops

The potential of hyperspectral measurements for early disease detection has been investigated by many experts over the last 5 years. One difficulties is obtaining enough data training and building a library. When goal to detect at previsible stage, before pathogen manifested either its first symptoms or in area surrounding existing symptoms, it impossible objectively delineate regions interest containing growth from areas without growth. To overcome this, we propose an image labelling segmentation algorithm that able (a) more label visible construction library (b) extend this pre-visible symptoms. This used create libraries late blight (Phytophthora infestans) potatoes two types leaf rust (Puccinia triticina Puccinia striiformis) wheat. model accuracies were compared between automatic classic visual delineation using logistic regression machine learning approach. modelling automatically labelled datasets higher than those manually ones both wheat, 98.80% P. infestans potato, 97.69% striiformis soft 96.66% durum