Multiscale Sensing For Disease Monitoring In Vineyard Production – MERIAVINO

The project aims to combine all the elements of sensing, processing, safety storage, and analysis to provide a novel reliable system for vine monitoring. Nowadays different sensor networks with low consumption are designed for agriculture providing answers to the challenges of sustainability through sensor technologies and automatic information analysis. In contrast to existing approaches, the MERIAVINO project proposes a multi-scale design of measurements and their integration (from ground to sky/space), supplementary to the classic IoT architecture, with printable sensors with low-energy. New AI methods for data fusion and modelling is developed. The project also includes data processing under anonymisation to meet protection requirements. The overall study and experimentation is carried out on a continental scale, where the agronomic parameters and climatic conditions vary.

- The experimental plots were selected in three European regions: Murfatlar - Romania, Corinthia - Greece and Amboise - France. We monitored the climate, tracked the development of the pathogenic fungus in relation to specific climatic conditions and the vine's stage of development, and determined the main physiological factors (stomatal conductance and relative chlorophyll content) in relation to the state of health of the vine and the environment.

- The environment ontology is used to describe the environmental conditions: atmospheric temperature, illumination level (solar radiation), humidity, atmospheric pressure, wind, etc. The device/network ontology is used to describe the details of the sensing and actuating devices of the system. The experimental system consists of two sensor kits, one standard and one designed and tested in the laboratory. The first step of the project was to identify the parameters that are monitored for vine disease identification. Sensors dedicated to the measurement of the parameters are identified. Sensors and IoT systems were acquired and implemented them in the experimental vineyards in France, Romania, and Greece.

- To integrate security into the solutions developed by the project, we focused on three different aspects: (1) technological watch to ensure project production reach actual security standards, (2) auditing and analyzing the security of solutions chosen or developed by the project (3) integrate security into the architecture of the developed solutions. We have chosen to rely on off-the-shelf solutions for communications between storage and acquisition for the initial deployment; this allows for a quick validation of both our platform and our general approach. The security analysis of these solutions shows that they have good security properties, and that it is not necessary to add a security mechanism for a first prototype.

- A state of the art of the method have been studied found very little research have addressed the problem of fusion of heterogeneous data fusion for disease detection and prediction. We developed a new method that combine 2D satellite images with the weather data. The results are promising, further development is underway for drone images. The correlations between the data from IoT sensors give the information about the state of the vine, different algorithms are tested to identify a prediction method to disease prevention.

At the end of the project we expected :

- Establishing a centralized cloud server to aggregate data from diverse vineyard plots across different countries. This centralized repository ensures accessibility and the ability to integrate data from various sources.

- Leverage three years of historical data to understand and model the impact of local climate behavior on vineyard health. This period offers the depth needed to identify trends and patterns of environmental influence.

- Real-time prediction using deep learning models with time series. This gives the ability to provide timely information, enabling appropriate decision-making and adaptation.

- Fusion of heterogeneous data, time series and multispectral/hyperspectral images. This combination of different data sources can be used to detect diseases and possibly detect their early appearance.

- The integration of federated learning techniques aligns well with the project's architecture, as it involves multiple sites generating data, and addresses privacy protection concerns while encouraging collaborative efforts. The goal is to aggregate machine learning models trained on data from various sources without sharing the raw data, ensuring data privacy and security.

- A practical demonstration to showcase the viability of the proposed solutions, enabling stakeholders to witness the tangible benefits of the project.

Simona Ghi?a, Mihaela Hnatiuc, Aurora Ranca, Victoria Artem, Madalina-Andreea Ciocan, 2022. Studies on the short-term effects of the cease of pesticides use on vineyard microbiome. IntechOpen in the book under the working title «Vegetation Dynamics, Changing Ecosystems and Human Responsibility«, ISBN 978-1-80356-138-7.

Mihaela Hnatiuc, Mirel Paun, Daniel Kapsamun IoT Sensors System for Vineyard Monitoring, 2022 11th International Conference on Frontiers of Intelligent Technologys, 2022

Mihaela Hnatiuc, Bogdan Hnatiuc, Sorin Sintea, Simona Ghita, Aurora Ranca, Victoria Artem, Bogdan Cristian Savin, IOT technology used in vineyard monitoring, 2021 IEEE 27th International Symposium for Design and Technology in Electronic Packaging (SIITME)

Mihaela Hnatiuc,Bogdan Hnatiuc, Aurora Ranca, Sorin Sintea, Victoria Artem, Simona Ghita, The methods for vine disease identification, 2021 IEEE 27th International Symposium for Design and Technology in Electronic Packaging (SIITME), DOI: 10.1109/SIITME53254.2021.9663713, Electronic ISSN: 2642-7036, ieeexplore.ieee.org/document/9663649

Computer Vision, IoT and Data Fusion for Crop Disease Detection Using Machine Learning: A Survey and Ongoing Research, Maryam Ouhami , Adel Hafiane , Youssef Es-Saady , Mohamed El Hajji and Raphael Canals , Remote Sens. 2021, 13(13), 2486.

Maryam Ouhami, Youssef Es-Saady, Mohammed El Hajji, Raphael Canals and Adel Hafiane “Meteorological Data and UAV Images for the Detection and Identification of GrapevineDisease Using Deep Learning”, IEEE E-HEALTH AND BIOENGINEERING CONFERENCE – EHB, 2022 10-th edition, Ia?i, Romania, November 17-19, 2022

Arun Pandian; V. Dhilip Kumar; Oana Geman; Mihaela Hnatiuc; Muhammad Arif; K. Kanchanadevi, « Plant Disease Detection Using Deep Convolutional Neural Network«, Applied Sciences, MDPI 2022-07-10

William Maillet, Maryam Ouhami, and Adel Hafiane. Fusion of satellite images and weather data with transformer networks for downy mildew disease detection. IEEE Access, 11 :5406–5416, 2022

MERIAVINO project advocates a multidisciplinary approach, which is based on several scientific fields to address the problem of disease and yield estimation in vineyard. The proposed methodology consists of inter-combining and implementing IoT, remote sensing and big data with a multi-scale approach in order to interconnect the vineyard parcels, as well as to develop a non-invasive, eco-friendly and low-cost technology for vine disease detection/warning. In order to reduce economic loss of both quantity and quality, and the environmental impact, various sensors, data fusion techniques and artificial intelligence and machine learning methods are combined along with the development of reprintable sensors for effective vineyard monitoring.The project results are then analysed and geo-visualised on compatible MobApp for end-users for decision-making and early prevention.