AGIcam: An Open-Source IoT-Based Camera System for Automated In-Field Phenotyping and Yield Prediction
A low-cost, solar-powered IoT platform for high-frequency crop monitoring and yield prediction in wheat breeding trials
IoT-based-Camera-Development/
├── 1_Camera_Development/ # Hardware design & 3D models
│ ├── 1_Enclosure_3DModel/ # STL files for 3D printing
│ └── 2_Program_on_RasPi/ # Raspberry Pi software
├── 2_Backend_System/ # Node-RED data pipeline
│ ├── flows.json # Complete data flow
│ └── data_transform.js # Data transformation
├── 3_AGIcam_Dashboard/ # Web interface
│ ├── index.html # Homepage
│ ├── dashboardlists.html # Sensor dashboards
│ └── planthealth.html # Image galleries
├── 4_Data_Analysis/ # ML models
│ ├── LSTM_TimeSerie_Yield.../ # LSTM implementation
│ └── RandomForest_TimeSerie_.../ # Random Forest models
└── requirements.txt # Python dependencies
AGIcam is an open-source Internet of Things (IoT) camera system designed for automated in-field phenotyping and yield prediction. Developed at Washington State University's Phenomics Lab, this platform enables continuous, high-frequency monitoring essential for capturing rapid phenological transitions and dynamic crop responses in breeding programs.
Figure 1: AGIcam sensor system deployed in a winter wheat breeding trial during the 2022 growing season
- Solar-Powered Autonomy: 6W solar panel with 6,400 mAh battery for season-long operation
- Wireless Connectivity: 4G LTE and Wi-Fi for real-time data transmission
- Dual Camera System: Synchronized RGB and NoIR imaging (3x daily capture-Adjustable depend on User's requirement)
- Edge Computing: On-device vegetation index calculation for 7 VIs
- Cloud Integration with MING Stack: Automated data transfer from Node-RED MQTT to InfluxDB with Grafana visualization
- Low Cost: $150-200 per sensor unit
Figure 2: System architecture of the AGIcam platform, illustrating its core components
The AGIcam platform consists of four main components:
- Hardware Development - Physical sensor design and 3D enclosures
- Backend System - Node-RED data pipeline and cloud integration
- Web Dashboard - Real-time monitoring interface
- Data Analysis - Machine learning models for yield prediction
- 18 sensors deployed across spring and winter wheat trials
- 85%+ uptime maintained throughout the growing season
- Sub-daily monitoring with 3 imaging sessions per day
- 7 vegetation indices computed in real-time
| Crop Type | Model | RMSE (kg/ha) | Error Rate |
|---|---|---|---|
| Spring Wheat | LSTM | 221.76 | 3.41% |
| Winter Wheat | LSTM | 210.28 | 1.62% |
| Spring Wheat | Random Forest | 544.79 | 8.60% |
| Winter Wheat | Random Forest | 1059.82 | 10.41% |
- Raspberry Pi Compute Module 3+ Lite
- Dual Raspberry Pi Camera V2 (RGB + NoIR)
- 6W Solar Panel + 6,400 mAh Battery
- Witty Pi 3 power management
- Custom 3D-printed enclosure
- OS: Raspbian Buster
- Backend: Node-RED, Python 3.7+
- Database: InfluxDB (time-series)
- Visualization: Grafana
- Cloud: Microsoft Azure with Bootstrap framework
Our research demonstrates AGIcam's effectiveness for:
- Yield prediction with LSTM achieving 1.62% error
- Phenological monitoring during critical growth stages
- High-throughput phenotyping in breeding programs
- Real-time decision support for crop management
If you use AGIcam in your research, please cite:
Paper:
Sangjan, W., Pukrongta, N., Buchanan, T., Carter, A. H., Pumphrey, M. O., & Sankaran, S. (2026).
AGIcam: An open-source IoT-based camera system for automated in-field phenotyping and yield prediction.
bioRxiv, 2026.01.13.699185. https://doi.org/10.64898/2026.01.13.699185
Dataset:
Sangjan, W., Pukrongta, N., Buchanan, T., Carter, A. H., Pumphrey, M. O., & Sankaran, S. (2025).
AGIcam Dataset: In-Field IoT Sensor Data for Wheat Phenotyping and Yield Prediction [Data set].
Zenodo. https://doi.org/10.5281/zenodo.17970104
This research was funded by:
- USDA-NIFA Competitive Project (Accession #1028108)
- Washington State University Hatch Project (Accession #1014919)
- WSU College of AHNRS Emerging Research Issues Grant (ERI-20-04)
- Lead Researcher: Worasit Sangjan - [email protected]
- Principal Investigator: Dr. Sindhuja Sankaran - [email protected]
- Institution: Washington State University, Phenomics Lab
This project is open source under the MIT License. See the LICENSE file for details.
© 2022 AGIcam - Phenomics Lab|Washington State University