Smart Mushroom Farming : integrating IoT and image processing for enhanced cultivation
Raghavan, Parthsarthi (2024)
Julkaisun pysyvä osoite on
https://urn.fi/URN:NBN:fi:amk-2024120231917
https://urn.fi/URN:NBN:fi:amk-2024120231917
Tiivistelmä
This thesis presents an IoT and image processing system to automate mushroom cultivation, addressing labor-intensive and inefficient traditional methods. A Raspberry Pi 4B serves as the central controller, integrating sensors (DHT22 for temperature and humidity, MH-Z19 for CO₂, BH1750 for light intensity) and actuators (fans, humidifier, UV bulb) to maintain optimal environmental conditions. YOLOv5 detects mushrooms in high-resolution images, while MobileNetV2 classifies growth stages. Real-time data is transmitted via MQTT to MongoDB Cloud, creating a feedback loop for automated adjustments.
Tested in a controlled acrylic box (590x396x245 mm), the system achieved
±1.8% monitoring accuracy, 99.8% uptime, and environmental adjustments
within 800 milliseconds. YOLOv5 delivered a mean average precision (mAP) of 0.758, and MobileNetV2 achieved 92% classification accuracy. Water and energy use decreased by 30% and 20%, respectively, compared to traditional methods. The modular design supports scalability and application to crops like tomatoes. Future work focuses on expanding datasets, deploying edge AI, and enhancing hardware for broader precision agriculture applications.
Tested in a controlled acrylic box (590x396x245 mm), the system achieved
±1.8% monitoring accuracy, 99.8% uptime, and environmental adjustments
within 800 milliseconds. YOLOv5 delivered a mean average precision (mAP) of 0.758, and MobileNetV2 achieved 92% classification accuracy. Water and energy use decreased by 30% and 20%, respectively, compared to traditional methods. The modular design supports scalability and application to crops like tomatoes. Future work focuses on expanding datasets, deploying edge AI, and enhancing hardware for broader precision agriculture applications.