Predicting wheat yield in agricultural industry using deep learning techniques: a review

Abstract

In the post-pandemic future, technology in the agriculture industry can improve food sustainability while moderating the use of resources of nature in a variety of conditions. Robotic tools for agriculture have been developed for crop planting, nursing, clearing weeds, pest management, and harvesting. The key aspects of crop growth and innovative agricultural engineering help farmers maximize crop yield. In the present investigation, it has been found that deep learning (DL) algorithms are used to enhance the predictability of wheat crop yield. The assessment and forecasting of wheat crop yields can be done with precision and dependability using satellite imagery. The scientific investigations in this study to predict wheat crop yield considered distinct factors, including various vegetation indices with remotely sensed imaging, climate-related conditions, nutrients in the soil, wheat plant diseases, and water scheduling. This study expounds a variety of DL strategies for predicting wheat production and found that many publications make use of long short-term memory (LSTM), along with residual network (ResNet) and deep neural networks (DNN). The performance measures, commonly harnessed in publications, are highlighted in this study, including coefficient of determination (R²), root mean square error (RMSE) and accuracy. This systematic evaluation of the literature on the wheat crop will open possibilities for future research for scholars.