Developing an Intelligent Agricultural System Based on Long Short-Term Memory

Abstract

Today, the agriculture industry has been developing intellectualization and automation proactively for reducing labor force and increase yields. In the past, farmers usually followed the rule of thumb to grow crops; however, due to the dramatic climate change, it becomes harder for farmers to cope with it by merely following the rule of thumb, which leads to crop damage. Therefore, it is vital to input scientific data development and technology for optimizing the environment parameters of crops and further enhance the yields. Additionally, many farms need to spread pesticides to avoid pests and diseases; yet, too much pesticide may cause soil alkalization. To enrich the growing-power of the lands, farmers will fertilize the lands; nevertheless, too much of it will also cause soil acidification, which will need to leave the land fallow to improve the soil quality. The study provides an intelligent agriculture system based on LSTM. The system develops an Internet of Things (IoT) to monitor the environmental conditions of soil, sunlight, and temperature; additionally, the research combines the information from the Central Weather Bureau for predicting the timing for watering and notifying farmers about the suggested amount of pesticides and fertilizers. The features of this article are as follows: 1. Build a clustering tree of crops according to the adaptability; 2. Calculate the critical values of each selected crop; 3. Develop an LSTM system that provides analyses according to the current soil conditions and weather forecast information; the system will reveal the conditions of the soil, and water the land to balance the condition and reach an optimal status if the soil pH is too high. The research is capable of enhancing crop yields and optimizing the land.