VELOS: Intelligent Pesticide and Irrigation Management in Precision Agriculture

VELOS represents a comprehensive smart ecosystem for precision agriculture, specifically designed for pest management and irrigation optimization in bean cultivation within the Prespa Region of Greece. This innovative project integrates cutting-edge technologies to revolutionize agricultural decision-making processes.

Project Overview

VELOS (Intelligent Pesticide and Irrigation Management in Precision Agriculture) is an open-source, modular, and scalable framework that leverages Internet of Things (IoT) technologies, Unmanned Aerial and Ground Vehicles (UAVs/UGVs), Low-Power Wide-Area Networks (LPWANs), Artificial Intelligence (AI), and Machine Learning (ML) techniques to extract knowledge and create integrated solutions for effective agricultural management.

System Architecture

The VELOS ecosystem follows an N-level architecture design to ensure flexibility, robustness, and efficiency while providing workload balancing across system units. The architecture comprises seven interconnected subsystems:

1. IoT Subsystem

• LoRaWAN Network: Utilizes Long Range Wide Area Network for flexible scalability and low development costs

• Telemetric Meteorological Stations: Seven stations distributed across the main bean growing area

• Real-time Data Collection: Continuous monitoring of soil moisture, temperature, humidity levels

• Cloud Integration: Data transmission to cloud-based servers using ADCON addVANTAGE software

1. UAV Subsystem

• Fleet Management: UAV-as-a-service model for coordinated aerial operations

• Mission Planning: Automated mission initiation, definition, and UAV assignment

• Flight Generation: Optimized flight paths for comprehensive crop monitoring

• Ground Control System: Centralized control and coordination of UAV operations

1. UGV Subsystem

• Custom Robotic Platform: Purpose-built Unmanned Ground Vehicle for precision agriculture

• Advanced Equipment: DC motors, robotic arm with spectral camera, obstacle avoidance sensors, GPS

• Optimal Path Finding: AI-driven algorithms considering energy consumption and mission time

• Enhanced Data Quality: Ground-level data collection to improve pest prediction accuracy

1. Pest Risk Threshold Subsystem

• Degree-Day Thresholds: Empirical prognostic models for seasonal pest occurrence prediction

• Target Pests: Helicoverpa armigera, Thrips sp., and Tetranychus urticae

• Disease Risk Indices: Epidemiological indicators for Uromyces phaseoli (bean rust)

• Validation Framework: Two-season data collection (2021-2022) for threshold validation

1. Pest Damage Detection Engine (PDDE)

• Multi-Model Approach: Portfolio of state-of-the-art CNN-based detection models

• Advanced Architectures: Faster-RCNN, SSD, RetinaNet, EfficientDet, YOLOv4, YOLOv5

• Image Processing: Comprehensive preprocessing including resize, augmentation, and denoising

• Damage Classification: Automated detection of arthropod pest damage and disease symptoms

1. Irrigation Forecasting Engine

• Regression Analysis: Multiple ML algorithms including SVMs, decision trees, random forest, MLP

• Data Integration: IoT sensor data combined with meteorological forecasts

• Preprocessing Pipeline: Advanced techniques for handling missing values and outliers

• Predictive Modeling: Accurate irrigation needs forecasting for optimal water management

1. VELOS Intelligent Decision-Making System (DSS)

• System Orchestration: Central coordination of all subsystems

• Three-Stage Pest Prediction: Integrated approach combining thresholds, UAV data, and UGV validation

• Recommendation Engine: Informed pesticide application and irrigation scheduling

• False-Positive Minimization: Multi-source validation to improve prediction accuracy

Experimental Setup and Validation

Field Network

• Four Pilot Fields: 4-7 acres each in the Prespa National Park area

• Cultivation Types: Two conventional and two organic plots for comparative analysis

• Monitoring Protocol: Bi-weekly observations throughout growing seasons

• Data Collection: Sequential pest monitoring and meteorological data gathering

Meteorological Infrastructure

• Seven Weather Stations: Strategically distributed across the study area

• Real-time Monitoring: Continuous temperature, humidity, and precipitation tracking

• Cloud Integration: Remote data transmission for immediate analysis

• Historical Data: Multi-season datasets for threshold development

Research Contributions and Publications

This project resulted in three significant scientific publications:

1. "Machine learning and deep learning for plant disease classification and detection"

Authors: V Balafas, E Karantoumanis, M Louta, N Ploskas Publication: IEEE Access 11, 114352-114377 Focus: Comprehensive review and analysis of ML/DL techniques for plant disease detection

1. "Intelligent Pesticide and Irrigation Management in Precision Agriculture: The Case of VELOS Project"

Authors: MD Louta, F Papathanasiou, P Damos, N Ploskas, M Dasygenis, et al. Publication: HAICTA, 91-99 Focus: Complete system architecture and implementation methodology

1. "Real-time disease detection on bean leaves from a small image dataset using data augmentation and deep learning methods"

Authors: E Karantoumanis, V Balafas, M Louta, N Ploskas Publication: Soft Computing, 1-13 Focus: Novel approaches for handling limited datasets in agricultural AI applications

Technical Innovation

Machine Learning Pipeline

• Convolutional Neural Networks: State-of-the-art architectures for image analysis

• Data Augmentation: Advanced techniques to overcome limited dataset challenges

• Multi-Model Ensemble: Portfolio approach for improved prediction accuracy

• Real-time Processing: Immediate analysis and decision-making capabilities

Integration Challenges

• Interoperability: Seamless communication between diverse subsystems

• Scalability: Modular design allowing easy expansion and modification

• Robustness: Fault-tolerant architecture ensuring continuous operation

• User Interface: Farmer-friendly interfaces for practical implementation

Environmental and Economic Impact

Sustainability Benefits

• Pesticide Reduction: Optimized application reducing environmental footprint

• Water Conservation: Precision irrigation minimizing resource waste

• Crop Yield Optimization: Improved quality and quantity through informed management

• Cost Reduction: Efficient resource utilization lowering production costs

Precision Agriculture Advancement

• Technology Integration: Successful combination of IoT, AI, and robotics

• Knowledge Generation: Data-driven insights for agricultural decision-making

• Farmer Empowerment: Advanced tools accessible to agricultural practitioners

• Research Foundation: Platform for continued agricultural innovation

Future Implications

The VELOS project demonstrates the transformative potential of integrating emerging technologies in precision agriculture, providing a blueprint for sustainable farming practices that balance productivity, profitability, and environmental stewardship. The system's modular architecture and open-source approach facilitate adoption and adaptation across different agricultural contexts and crop types.