A Gated Recurrent Units (GRU)-Based Model for the Prediction of Soybean Sudden Death Syndrome with Time-Series Satellite Imagery

摘要

Early detection and timely management of plant diseases are essential for agricultural production. Traditional manual inspection is often time-consuming and laborious. Recently, automated imaging techniques have been successfully applied to the detection of plant diseases. However, these methods typically use one-time image. This paper presents a Gated Recurrent Units (GRU)-based model to predict the soybean sudden death syndrome (SDS) rates at multiple time periods. Different from plant disease recognition using near sensing imagery, the proposed method uses a satellite imagery dataset collected from PlanetScope as the training set. Instead of using static individual imagery, it converts the original imagery into time-series data. The experiment results show that the proposed method can improve the prediction accuracy by up to 10% with time-series prediction. The proposed method can also be applied predict SDS at a future time.