Automatic Detection of Water Stress in Corn Using Image Processing and Deep Learning

Water stress is one of the main environmental constraints that directly disrupts agriculture and global food supply, thus early and accurate detection of water stress is necessary in order to maintain high agricultural productivity. Using an image dataset collected during a dedicated experiment, we propose a new method for water stress level classification using deep learning and digital images only. Classification is performed in two stages, using a Convolutional Neural Network for spatial feature extraction and a Long Short-Term Memory for temporal features extraction. Outperforming all other methods examined, our model is able to classify five different levels of water stress with 91.7% accuracy and Mean Absolute Error of 0.1, and to detect changes in water stress levels during the day.