Deep Learning-Based Prediction of Plant Growth and Yield in Greenhouse Environments

Abstract

For greenhouse growers and farmers in general, accurate forecasts of plant development and productivity are critical. Growers can enhance environmental control, match supply with demand, and minimise costs by developing models that accurately forecast growth and output. Powerful new analytical tools may be gained from recent advances in Machine Learning and in particular, Deep Learning (DL). Tomato yield forecasting and Ficus benjamina stem development will be predicted using ML and DL approaches in a controlled greenhouse setting in the proposed project. In the prediction formulations, we use a novel deep recurrent neural network (RNN) based on the LSTM neuron model. The RNN architecture models the intended growth parameters using the previous yield, growth, and stem diameter measurements, as well as microclimate circumstances. Support vector regression and random forest regression are compared in a researchutilising the mean square error criteria in order to assess the effectiveness of the various approaches. Results from the EU Interreg SMARTGREEN project (2017-2021) in two greenhouses in Belgium and the UK have shown great promise, according to the statistics given.