Sensors substitution using AI for agriculture soil moisture monitoring

Tay, Kai Sheng (2025) Sensors substitution using AI for agriculture soil moisture monitoring. Final Year Project, UTAR.

PDF Download (8Mb)

Abstract

This project focuses on the growing trend of the Internet of Things (IoT) and Machine Learning (ML) in precision agriculture, specifically sensor substitution using AI for agriculture soil moisture monitoring. Traditional soil moisture sensors face challenges such as environmental degradation and maintenance costs, leading to the need for a more reliable and scalable solution. This project aims to develop an AI-powered soil moisture prediction system that enhances irrigation management by utilizing temperature and humidity data instead of direct soil moisture readings. The system consists of IoT hardware (ESP32 microcontroller and DHT22 sensor), a cloud-based web application, and a trained machine learning model. The collected sensor data is sent to a real-time monitoring dashboard, where users can view live data trends and change to developer mode when data collection is needed for new plants. The AI model which is trained using ensemble method which contains random forest regressor and gradient boosting regressor to process the collected information to predict soil moisture levels and detect anomalies, providing smart irrigation recommendations. The key novelty in this project is eliminating the need for direct soil moisture sensors through reliable AI estimation, integration of developer mode triggering for ESP32 via backend control and a modular dashboard design for visualizing data and database integration. The experimental results show promising accuracy in soil moisture predictions and support the efficient irrigation decision-making. The systems improve the scalability, maintainability and also cost-effectiveness in smart farming, contributing toward AI-driven agriculture and better plant monitoring management.

Actions (login required)