More information
Description
The GrapeGuard project, supported by the EEA and Norway Grants and led by Robotics.AI SRK, is dedicated to tackling the complex issue of vine diseases that afflict grapevines at various stages of growth, including downy mildew, oidium, gray mold, black rot, and excoriose. These diseases pose a significant threat to vineyards and can result in substantial crop losses if not detected promptly. Leveraging their expertise in robotics and AI, the project aims to develop early diagnosis solutions for these plant abnormalities, crucial for maintaining the health and productivity of vineyards and ensuring a stable yearly wine production. This endeavor is vital in the face of changing climate patterns and unpredictable weather events affecting agriculture.
The project intends to access support for the fundamental R&D parts of the project. This step is important to ensure the cutting-edge innovative solutions are integrated in the GrapeGuard project. This project aims at development and validation of the potential low-altitude aerial images in a common topological map for vineyard state evaluation.
Achieving these targets require a variety of key activities, including:
- Project Management
- Training Activities
- Project Promotion
- The development of a new innovative map platform
Summary of project results
The GrapeGuard project aimed to address several significant challenges in the vineyard industry. One of the primary issues was the detection of vine diseases (VDD) such as downy mildew, oidium, gray mold, black rot, and excoriose, which can severely impact grapevine health and crop productivity. These diseases, if not identified early, can cause losses of up to 70-80% of the annual wine crop.
Additionally, the project sought to tackle the effects of climate change, unexpected precipitation patterns, and temperature fluctuations, which further complicate the detection and management of these diseases. Traditional visual examination methods were insufficient for timely and accurate disease identification, necessitating the development of more advanced solutions.
The project also aimed to bridge the gap between early-stage research focusing on aerial observation and the need for close proximity observation available from ground-level robots. By integrating low-altitude aerial images with ground-level data, the project intended to create a comprehensive topological map for vineyard state evaluation, enhancing early-stage disease detection and providing real-time information on large-scale productions.
The project undertook several key activities to address the challenges in vineyard disease detection and management. The project aimed to develop and validate low-altitude aerial imaging integrated with ground-level data to create a comprehensive topological map for vineyard state evaluation. This approach was intended to enhance early-stage disease detection and provide real-time information on large-scale productions.
The project was structured around three main objectives:
Close-up UAV data acquisition and georeferencing: This involved integrating low-level flight UAVs in an autonomous waypoint mapping application. The team investigated various flight modes, speeds, and altitude configurations to optimize image quality under different light and weather conditions. The goal was to ensure high-quality images for the image processing pipelines used in disease detection.
Early-stage detection with enhanced deep networks on the server-side: Using transfer learning tools from NVidia, the project focused on developing custom-trained network architectures for VDD detection. This included manual data labeling supported by local viticulture specialists and data augmentation to improve the detection accuracy.
Development of a minimum viable product (MVP) for real farming conditions: The project aimed to validate the proposed services in real-life vineyard settings. This involved on-site specification, solution customization, and client-side testing to ensure the technology met the practical needs of vineyard management.
The project produced several notable outputs, including:
- A georeferenced image database created with autonomous UAVs.
- An early-stage VDD detection system developed using NVidia''s transfer learning toolkit.
- A user-friendly application for customers to control and evaluate the system.
- Validation of the service in different vineyards, demonstrating the practical benefits of the technology.
The GrapeGuard project successfully tackled several critical challenges in the vineyard industry, particularly the early detection of vine diseases like downy mildew, oidium, gray mold, black rot, and excoriose. These diseases, if not identified early, can cause significant crop losses. The project developed an innovative system using UAVs and deep learning networks to provide real-time information on vineyard health, enabling timely interventions and reducing the risk of severe crop damage.
By integrating low-altitude aerial images with ground-level data, the project created a comprehensive topological map for vineyard state evaluation. This allowed vineyard managers to monitor and manage their crops more effectively, improving overall productivity and efficiency. Additionally, the early detection system helped reduce the need for preventive chemical treatments, leading to a decrease in pesticide use and promoting more sustainable agricultural practices.
The project also had significant economic and environmental impacts. It projected an annual growth in turnover of 15% by the end of 2026 and a substantial increase in net operational profit. The project also aimed to create new jobs and submit new Intellectual Property Rights applications, fostering innovation. Environmentally, the project implemented measures to reduce water pollution and fuel consumption, contributing to more eco-friendly vineyard management.