Smart Irrigation System Based on Feedback for Digital Agriculture

Digital agriculture is able to improve the convenience and productivity by digitalizing occurred event in agricultural process. The irrigation system is the most important element in agricultural process. There are various research on automation of irrigation system. Most of research have the disadvantage that administrator need to intervene to irrigate. In this work, we propose a smart irrigation system that can monitor agricultural environment and can decide irrigation period and irrigation time. Also, we design machine learning models base on time series data such as CNN, Simple RNN, LSTM to classified soil texture. Performance of the classification algorithm was evaluated using the confusion matrix, the classification performance was evaluated about 90%. In order to implement tiny machine learning on an embedded system in future work., we will consider Simple RNN that has the fewest parameters of them.