Serin et al. (2025) AI-Driven Smart Farming for Automated Plant Health Monitoring and Nutrient Deficiency Detection

Identification

• Journal: ASEAN Journal of Scientific and Technological Reports
• Year: 2025
• Date: 2025-12-14
• Authors: J. Serin, G.Gifta Jerith, Veemaraj Ebenezer, K.Arul Jeyaraj, A. Jenefa, M. Vargheese
• DOI: 10.55164/ajstr.v29i1.258175

Research Groups

• Department of Computer Science, Women’s Christian College, Chennai, Tamil Nadu, India
• Department of AI & ML, Malla Reddy University, Hyderabad, India
• Division of Data Science and Cyber Security, Karunya Institute of Technology and Sciences, Coimbatore, Tamil Nadu, India
• Department of Electronics and Communication Engineering, PSNA College of Engineering and Technology, Dindigul, Tamil Nadu, India
• Department of Computer Science and Engineering, PSN College of Engineering and Technology, Tirunelveli, Tamil Nadu, India

Short Summary

This paper presents an AI-driven Internet of Things (IoT) system for automated plant health monitoring and early nutrient deficiency detection, integrating multi-sensor data with camera-based leaf analysis. The system achieved 89.0% accuracy in classifying plant health and nutrient deficiencies using a DenseNet 121 model, enabling autonomous irrigation and localized cooling.

Objective

• To develop an integrated Internet of Things (IoT) and Artificial Intelligence (AI) pipeline for automated plant health monitoring and early nutrient deficiency detection, coupled with autonomous environmental control for small plots and greenhouses.

Study Configuration

• Spatial Scale: Small plots and greenhouse environments.
• Temporal Scale: Continuous, near real-time data-driven monitoring and response.

Methodology and Data

• Models used: DenseNet 121 (for leaf image classification), MobileNet V2 (baseline for comparison), ATmega microcontroller (for system control and actuation).
• Data sources:
  • RGB leaf images (224 × 224 pixels) captured in field conditions and labeled by experts into healthy, nitrogen deficiency, and phosphorus deficiency classes.
  • Soil moisture sensor.
  • DHT sensor (temperature and humidity).
  • Dataset: 1,200 images (480 Healthy, 360 Nitrogen deficiency, 360 Phosphorus deficiency) split into 70/15/15 train/validation/test sets plot-wise.

Main Results

• The DenseNet 121 classifier achieved an accuracy of 89.0% on a held-out test set for detecting nutrient deficiencies.
• This performance surpassed a MobileNet V2 baseline, which achieved 82.0% accuracy under identical training conditions, representing a 7.0 percentage point absolute gain and a 38.9% reduction in classification error.
• Prototype deployments demonstrated reduced manual checks and improved response to moisture and heat stress through closed-loop irrigation and localized cooling.
• The integrated IoT and AI pipeline proved practical for early detection of nutrient deficiencies and autonomous actuation in small plots and greenhouses.

Contributions

• Development of a comprehensive IoT and AI system that fuses soil moisture, temperature, and humidity sensing with camera-based leaf analysis for automated plant health monitoring.
• Implementation of a deep learning model (DenseNet 121) that significantly outperforms a MobileNet V2 baseline in nutrient deficiency detection accuracy.
• Creation of a closed-loop control system with a microcontroller for autonomous irrigation and localized cooling based on real-time sensor data and AI insights.
• Demonstration of a practical and effective solution for reducing manual labor, optimizing resource use (water, power), and improving crop health in small-scale agricultural settings.

Funding

• This research received no external funding.

Citation

@article{Serin2025AIDriven,
  author = {Serin, J. and Jerith, G.Gifta and Ebenezer, Veemaraj and Jeyaraj, K.Arul and Jenefa, A. and Vargheese, M.},
  title = {AI-Driven Smart Farming for Automated Plant Health Monitoring and Nutrient Deficiency Detection},
  journal = {ASEAN Journal of Scientific and Technological Reports},
  year = {2025},
  doi = {10.55164/ajstr.v29i1.258175},
  url = {https://doi.org/10.55164/ajstr.v29i1.258175}
}