Periodic Reporting for period 1 - CrackSense (High throughput real-time monitoring and prediction of fruit cracking by utilising and upscaling sensing and digital data technologies)
Período documentado: 2023-01-01 hasta 2024-06-30
Smart Farming applies modern Information and Communication Technologies (ICT) into agriculture, and it is based on sensor data feedback. This proposal aims to develop and upscale sensing technologies to provide real time sensor data for addressing the problem of fruit cracking.
Fruit cracking is common in plantations, and may cause large scale yield loss. Its intensity is affected by intrinsic plant traits, by environmental parameters, and by management practices. Of the environmental and horticultural variables, climate and irrigation are considered as major players in determining cracking intensity, respectively. Once in a few years the disorder aggravates to more than 50 % of the fruits. It is hypothesised that extreme climatic conditions and sub-optimal irrigation regimes at certain phenological stages, reduce peel resistance to growth strains. There is no comprehensive model predicting climatic conditions that promote cracking incidence and severity – for specific crops and specific locations. The interaction between the climatic variables, management and temporal fruit development is unknown. Therefore, cracking is the combined effect of multiple factors, environmental and endogenous, and has an erratic and unpredicted nature.
Prediction of cracking incidence based on various sensing/imaging technologies by fruits and trees scanning as early as possible before the disorder becomes visible, could be ideal. So far, imaging technologies have not been developed, but upscaling and combining proximal and remote sensing tools might well allow cracking detection and the development of year-, plot- and region-based risk assessment models. Modelling requires, on the one hand, the collection of agri-environmental parameters for a given plot and growing region, and, on the other hand, high throughput and precise monitoring of fruit growth and cracking development, for as many fruits, plots and regions as possible. Earth Observation satellite data, meteorological stations and other sensors allow the collection of agri-environmental variables at all the above-mentioned levels. Other variables, such as plot location, topography, microclimate, soil texture, horticultural practices, and irrigation level, are also feasible features to collect at the required scale. Therefore, generation of large databases pertaining to different fruit species and containing all agri-environmental variables is a feasible. In contrast, collection of high throughput cracking data at the level of the individual fruit, tree, and plot, and the exact timing of cracking, are challenging tasks, and are therefore limited to a small number of trees and plots. The present proposal aims to upscale sensing methods for detection of cracking at the fruit, tree and plot levels. Exploiting a number of remote and proximal sensing tools is a key to generate the required dataset that could be combined with other ancillary features and be fed into prediction models. Furthermore, ML and AI methods could reveal the complex relationship between the various features and provide robust estimations of the risk of cracking. Upscaling crack monitoring by high throughput tools, in an automatic and timely manner, would improve our understanding of the phenomenon of cracking and eventually benefit the growers by providing them with better tools to manage cracking incidences and the subsequent yield loss.
Therefore, CrackSense key objective is to upscale sensing technologies to monitor fruit cracking and yield loss at the fruit, tree, plot and regional levels, and to integrate this data with agri-environmental monitoring data in order to generate models for predicting cracking incidence and risk at the fruit/plot/regional/country levels for a given year.
Fruit cracking is common in plantations, and may cause large scale yield loss. Its intensity is affected by intrinsic plant traits, by environmental parameters, and by management practices. Of the environmental and horticultural variables, climate and irrigation are considered as major players in determining cracking intensity, respectively. Once in a few years the disorder aggravates to more than 50 % of the fruits. It is hypothesised that extreme climatic conditions and sub-optimal irrigation regimes at certain phenological stages, reduce peel resistance to growth strains. There is no comprehensive model predicting climatic conditions that promote cracking incidence and severity – for specific crops and specific locations. The interaction between the climatic variables, management and temporal fruit development is unknown. Therefore, cracking is the combined effect of multiple factors, environmental and endogenous, and has an erratic and unpredicted nature.
Prediction of cracking incidence based on various sensing/imaging technologies by fruits and trees scanning as early as possible before the disorder becomes visible, could be ideal. So far, imaging technologies have not been developed, but upscaling and combining proximal and remote sensing tools might well allow cracking detection and the development of year-, plot- and region-based risk assessment models. Modelling requires, on the one hand, the collection of agri-environmental parameters for a given plot and growing region, and, on the other hand, high throughput and precise monitoring of fruit growth and cracking development, for as many fruits, plots and regions as possible. Earth Observation satellite data, meteorological stations and other sensors allow the collection of agri-environmental variables at all the above-mentioned levels. Other variables, such as plot location, topography, microclimate, soil texture, horticultural practices, and irrigation level, are also feasible features to collect at the required scale. Therefore, generation of large databases pertaining to different fruit species and containing all agri-environmental variables is a feasible. In contrast, collection of high throughput cracking data at the level of the individual fruit, tree, and plot, and the exact timing of cracking, are challenging tasks, and are therefore limited to a small number of trees and plots. The present proposal aims to upscale sensing methods for detection of cracking at the fruit, tree and plot levels. Exploiting a number of remote and proximal sensing tools is a key to generate the required dataset that could be combined with other ancillary features and be fed into prediction models. Furthermore, ML and AI methods could reveal the complex relationship between the various features and provide robust estimations of the risk of cracking. Upscaling crack monitoring by high throughput tools, in an automatic and timely manner, would improve our understanding of the phenomenon of cracking and eventually benefit the growers by providing them with better tools to manage cracking incidences and the subsequent yield loss.
Therefore, CrackSense key objective is to upscale sensing technologies to monitor fruit cracking and yield loss at the fruit, tree, plot and regional levels, and to integrate this data with agri-environmental monitoring data in order to generate models for predicting cracking incidence and risk at the fruit/plot/regional/country levels for a given year.
Major achievements include:
1.The project developed a comprehensive Data Management Plan aimed at coordinating the scientific activity.
2. Proximal sensing of fruits and wetness accumulation on them was initiated in WP2. A data acquisition platform - TOMMY - was designed in Germany and replicated in Greece, France and Israel. The system has been calibrated, and operated in all four countries. The data has been shared among partners, and an open access database has already been created, and published.
3. Fruit images were acquired and first algorithms were developed, for early detection of fruit cracks and risk of cracking.
4. From the beginning of the project, WP3 concentrated on establishment of the infrastructure to collect the data upon which the decision support systems will be based, along with exploration of agronomical processes that can potentially mitigate the risk of cracking. Experimental plots for all four target crops have been established across nine geographical regions in four countries (Greece, France, Germany, and Israel). To generate a wide range of cracking risk levels, experimental plots undergo various treatments, including irrigation management, application of plant growth regulators, and others. A wide range of measurement technologies are employed across all experimental sites to generate comprehensive datasets characterizing the digital signature of each crop. This includes data collected from local sensors, manual in-field measurements, UAV and satellite imagery, as well as precise yield and cracking incidence data.
5. Sensor and other data from experimental plots were used to train a model correlating between tree physiological status and cracking intensity. The model shows good correlation between citrus and pomegranate canopy area and cracking intensity (Israel), providing proof of concept that sensing tools could be used to access cracking intensity. The model will be implemented for all other experimental plots
5. In WP4, a central repository was created, for data from experimental sites and regional data collection. The data are used to monitor and predict fruit cracking on different scales (fruit, tree, field, and regional scale), thereby enabling efficient agricultural resource management. Data collection protocols were finalized, a process that resulted in the creation of a database engine with MariaDB infrastructure, mostly integrating sensor data through an administration panel accessible under the authorization for specific partners. Towards progressing the design of the CrackSense web-based platform, comprehensive system requirements were provided to guide the initial version of the CrackSense Platform. The platform will primarily host the Decision Support System. An initial set of mockups was created for the platform's user-friendly interface, specifying technical details encompassing functionalities.
The scientific activities performed during the reporting periods (WP2, WP3 and WP4) are highlighted using transparent box in the attached image describing CrackSense project.
1.The project developed a comprehensive Data Management Plan aimed at coordinating the scientific activity.
2. Proximal sensing of fruits and wetness accumulation on them was initiated in WP2. A data acquisition platform - TOMMY - was designed in Germany and replicated in Greece, France and Israel. The system has been calibrated, and operated in all four countries. The data has been shared among partners, and an open access database has already been created, and published.
3. Fruit images were acquired and first algorithms were developed, for early detection of fruit cracks and risk of cracking.
4. From the beginning of the project, WP3 concentrated on establishment of the infrastructure to collect the data upon which the decision support systems will be based, along with exploration of agronomical processes that can potentially mitigate the risk of cracking. Experimental plots for all four target crops have been established across nine geographical regions in four countries (Greece, France, Germany, and Israel). To generate a wide range of cracking risk levels, experimental plots undergo various treatments, including irrigation management, application of plant growth regulators, and others. A wide range of measurement technologies are employed across all experimental sites to generate comprehensive datasets characterizing the digital signature of each crop. This includes data collected from local sensors, manual in-field measurements, UAV and satellite imagery, as well as precise yield and cracking incidence data.
5. Sensor and other data from experimental plots were used to train a model correlating between tree physiological status and cracking intensity. The model shows good correlation between citrus and pomegranate canopy area and cracking intensity (Israel), providing proof of concept that sensing tools could be used to access cracking intensity. The model will be implemented for all other experimental plots
5. In WP4, a central repository was created, for data from experimental sites and regional data collection. The data are used to monitor and predict fruit cracking on different scales (fruit, tree, field, and regional scale), thereby enabling efficient agricultural resource management. Data collection protocols were finalized, a process that resulted in the creation of a database engine with MariaDB infrastructure, mostly integrating sensor data through an administration panel accessible under the authorization for specific partners. Towards progressing the design of the CrackSense web-based platform, comprehensive system requirements were provided to guide the initial version of the CrackSense Platform. The platform will primarily host the Decision Support System. An initial set of mockups was created for the platform's user-friendly interface, specifying technical details encompassing functionalities.
The scientific activities performed during the reporting periods (WP2, WP3 and WP4) are highlighted using transparent box in the attached image describing CrackSense project.