Precision agriculture: exploration of deep learning models for farmland mapping

Abstract

Precision is required for agricultural advancements to be sustainable. Traditional farming lacks effective monitoring, resulting in resource waste and environmental problems. Farmland mapping is important for agricultural management and land-use planning. The use of deep learning techniques in farmland mapping is increasing rapidly. Excellent results have been generated from deep learning approaches in a number of applications, such as image processing and prediction. Agricultural agencies are now considering different applications of deep learning including land mapping, crop classification, and monitoring of paddy fields. This paper shall explore different deep learning models that are commonly used for image processing specifically in land mapping. The three deep learning models convolutional neural network (CNN), long short-term memory (LSTM), and recurrent neural network (RNN) were evaluated to find out which among the deep learning models is best for land mapping. It compares the classification accuracy of the models on image processing and it can be concluded that CNN algorithm normally makes better results when compared to other deep learning models. This study offers guideline and suggestions to researchers who are interested in contributing to the field of precision agriculture with the used of deep learning techniques.