Soil Nutrient Deficiency Detection of Lime Trees using Signal-based Deep Learning

This research work is initiated to create another alternative for the small farmers to identify crop nutrient deficiency by using modern technology that is more efficient and more precise than the conventional method of using eyes and hands. Most of the current plant health classification techniques focus on image-based learning methods. The presented work employs signal-based techniques to identify plant nutrient deficiencies. The classification process uses datasets retrieved via Xiaomi plant sensor that measures soil moisture and fertility, surrounding light, and ambient temperature of three lime trees. Each plant has different conditions that represents healthy plant, and sick plants (one is less fertilizer, another is less water). The objective of this project is to detect plant nutrient deficiency using signal-based deep learning methods. A comparison of based deep learning methods between RNN (Recurrent Neural Network) and MLP (Multilayer Perceptron) is performed using Python. Based on the findings of this experiment, MLP performs better with maximum accuracy of 98.6% using softmax activation function while RNN maximum accuracy is 95.4% using sigmoid activation function.