Estimation of Water Stress in Potato Plants Using Hyperspectral Imagery and Machine Learning Algorithms

This work presents quantitative detection of water stress and estimation the level: none, light, moderate, severe on potato crops. We use hyperspectral imagery state art machine learning algorithms: random decision forest, multilayer perceptron, convolutional neural networks, support vector machines, extreme gradient boost, AdaBoost. The in crops is carried out two different phenological stages plants: tubers differentiation maximum tuberization. algorithms are trained with a small subset each image corresponding to plant canopy. results improved using majority voting classify all canopy pixels images. indicate that both level can be obtained good accuracy, further by voting. importance band images classification assessed forest which perform best overall