TLDR: This research details the development of a smart aeroponic greenhouse that uses IoT for environmental control and AI for plant disease detection. The system continuously monitors conditions like temperature and humidity, adjusting them for optimal Geranium plant growth. An AI framework, primarily using the VGG-19 algorithm, was developed to detect drought stress and rust leaves, achieving up to 92% accuracy. This integration allows for remote monitoring, automated irrigation, and early disease diagnosis, significantly improving greenhouse management and crop production efficiency.
In an effort to address critical challenges in crop production, such as water scarcity, land limitations, and the need for effective plant disease control, researchers have developed an innovative smart aeroponic greenhouse. This experimental setup integrates the Internet of Things (IoT) and artificial intelligence (AI) to continuously monitor environmental conditions and the health of Geranium plants, aiming to optimize growth and detect diseases early.
Traditional agricultural methods face increasing pressure to meet global food demands while minimizing resource consumption. Soilless cultivation techniques, particularly aeroponics, offer significant advantages by requiring less space and water compared to conventional farming. Aeroponic systems allow plant roots to grow in a biphasic environment of liquid and air, promoting better oxygen contact and nutrient uptake. These systems can lead to higher yields, year-round production, and substantial savings in water, nutrients, and pesticides.
The Need for Smart Monitoring
While aeroponic systems are beneficial, achieving optimal productivity requires precise, instantaneous monitoring of environmental factors like temperature, humidity, oxygen, carbon dioxide, light, and nutrient levels. Manual monitoring is time-consuming, costly, and prone to human error, often leading to coarse data resolution and delayed responses to critical changes. To overcome these limitations, the researchers implemented an IoT-based system for remote and autonomous monitoring.
Integrating IoT for Environmental Control
The smart greenhouse, spanning 9 square meters, was designed with controlled artificial lighting, an electric heating system, and a cellulose ventilation cooling system. Ultrasonic humidifiers were used to maintain optimal humidity. Multiple temperature-humidity sensors were strategically placed to feed data to a central processing unit, which then activated or deactivated the heating, cooling, humidification, and ventilation systems as needed. This data is also published online to users via the Ubidots platform, allowing for remote oversight. For more details on the technical implementation, you can refer to the original research paper here.
The aeroponic feeding system uses polymer cultivation boxes with centrifugal nozzles to deliver nutrient solutions to the plant roots. Excess solution is collected and recirculated, conserving water and nutrients. Ultraviolet lamps are also integrated into the storage tanks to disinfect the nutrient solution and prevent contamination.
AI for Early Disease Detection
Beyond environmental control, a crucial aspect of this research was the development of an AI-based framework for plant disease detection. The team focused on identifying drought stress and rust leaves in Geranium plants. They utilized transfer learning with pre-trained convolutional neural networks (CNNs): VGG-19, InceptionResNetV2, and InceptionV3.
A dataset of 1,000 Geranium leaf images from industrial greenhouses, categorized into healthy, drought stress, and rust leaves, was prepared and augmented to 5,000 images to enhance the training of the AI models. The models were trained, validated, and tested, with VGG-19 showing the highest accuracy in classifying these conditions.
Key Findings and Performance
Preliminary results demonstrated the effectiveness of the IoT system in continuously publishing environmental data such as temperature, humidity, water flow, and tank volume online. It also successfully adjusted controlled parameters to maintain an optimal growth environment.
For disease detection, the VGG-19 algorithm outperformed the others, achieving an impressive 92% accuracy in identifying drought stress and rust leaves from healthy leaves on the industrial greenhouse dataset. When tested on data from the experimental greenhouse, VGG-19 maintained the highest overall accuracy at 86.34%, with 94.44% accuracy for healthy leaves, 75.60% for drought stress, and 86.84% for rust leaves. While the accuracy for drought stress was slightly lower, the VGG-19 algorithm consistently showed superior performance across all categories compared to InceptionResNetV2 and InceptionV3.
Also Read:
- Satellite Data Reveals Hidden Threat to Tomato Crops
- Advanced Sensing and AI Uncover Hidden Threat to Tomato Crops
Future Directions
This research highlights the significant potential of integrating IoT and AI in aeroponic greenhouses to enhance crop production and health. The ability to remotely monitor conditions and instantly detect diseases in their early stages allows for timely interventions, reducing economic risks and improving overall efficiency. Future work aims to further increase the accuracy of the algorithms by training them on larger and more diverse datasets, exploring more plant varieties, and identifying a wider range of diseases.
