Greater commercial demands, stricter quality standards and margins increasingly sensitive to operational errors make data-driven management necessary. A team from the University of O’Higgins has developed and implemented precision agriculture technologies in sweet cherry cultivation.
The sweet cherry tree (Prunus avium L.) currently leads Chilean fresh fruit exports in terms of value. According to data from the Office of Agricultural Studies and Policies (ODEPA), in the cumulative period between September 2025 and January 2026 the sector generated US$2.465 billion FOB (approximately €2.268 billion), with around 541,000 tonnes shipped. These figures represent 66% of the value and 44% of the total volume of fresh fruit exported by the country. In addition, by the end of the season, shipments of this crop reached around 112 million boxes, equivalent to approximately 561,000 tonnes, with China accounting for about 87% of the exported volume.
This performance comes within a context of sustained growth of the crop in Chile, which currently reaches around 77,766 planted hectares. Over the past five years, the area has grown at an annual rate close to 14.9%, with a particularly important role played by the O’Higgins Region, which accounts for 26.1% of the national area. Although a decrease in unit prices was observed in the latest season, associated with greater supply and a slower start to demand in the Asian market, industry projections indicate that both planted area and exported volumes will continue to expand in the coming years.
Sweet cherry orchard management is highly complex. Each season requires critical decisions related to crop load, irrigation, nutrition, growth regulation, thinning and the definition of the optimal harvest time. Added to this is the marked variability within the orchard, both between trees and between different sectors of the farm, which generates significant differences in terms of vigor, fruit set, fruit size and ripening. At the same time, increasing climate variability raises production uncertainty. In a region such as O’Higgins, where microclimates can vary even between neighboring farms, these differences can translate into very different management strategies.
High-value fruit production today faces a scenario marked by greater commercial demands, stricter quality standards and margins increasingly sensitive to operational errors. In this context, production management ceases to be merely a matter of accumulated experience and becomes a challenge of integrating information across multiple scales. As a result, decisions based solely on visual observation or historical experience are increasingly insufficient in the face of more intensive production systems.
Given these complexities, the need emerges for data-driven management, in which precision agriculture, through the systematic collection of information, offers a strategy to optimize the quality of decision-making. Through sensors, images and processing algorithms, it is possible to quantify variables that in the past were assessed subjectively, such as the number of flowers, fruit load, the spatial distribution of fruit or canopy volume. The use of artificial intelligence (AI) and predictive models makes it possible to transform these data into actionable information, reducing uncertainty and improving efficiency in orchard management. In this way, growers can anticipate production scenarios and adapt their management decisions with a stronger technical foundation.
For this reason, an interdisciplinary group of academics from the Institute of Engineering Sciences (ICI) and the Institute of Agri-Food, Animal and Environmental Sciences (ICA3) of the University of O’Higgins (UOH), together with postgraduate students and professionals, has been working for several years on the development and implementation of precision agriculture technologies in sweet cherry cultivation. This line of applied research integrates data collection through sensors, computer vision, three-dimensional modeling of fruit and trees, and AI, tools that now guide the technology transfer promoted by the university to the fruit sector in the O’Higgins Region.
In sweet cherry cultivation, complexity increases due to the short reaction time in the field and the large volume of fruit that must be assessed during the season. Estimating crop load, defining pruning strategies, fine-tuning irrigation and synchronizing critical operations require timely decisions that often have to be made with partial information that is not always representative of the real variability within the orchard. Thousands of fruit per tree, microclimatic differences between plots and heterogeneous physiological responses mean that small deviations can translate into significant impacts on yield and fruit quality. In this context, the current challenge is not to replace existing agronomic knowledge, but to expand and structure it through tools capable of capturing, processing and transforming large volumes of data into useful information for decision-making.
The integration of enabling technologies, such as connected sensors (IoT), artificial intelligence models, robotic systems and advanced analytics, makes it possible to observe the crop at individual-tree level, identify patterns invisible to the naked eye and anticipate scenarios before they result in losses. In this way, the orchard begins to be approached as a dynamic and spatially differentiated system.
Historically, fruit counting has been carried out manually, a procedure that introduces variability into the results and requires a considerable investment of time and human resources. However, the number of fruit alone does not fully describe the production status of an orchard: aspects such as size, shape and caliber distribution within the tree are equally decisive for estimating production potential and planning management operations. The current challenge is not limited to improving counting accuracy, but aims to develop methods that allow fruit to be geometrically characterized in an objective, systematic and scalable way.
The development of AI- and computer vision-based techniques is opening up new possibilities for agricultural monitoring. These technologies allow computers to analyze images of trees and identify key structures, such as spurs, flowers or fruit. Based on this type of analysis, different monitoring and production assessment tasks can be automated. One of the most relevant applications is the estimation of fruit quantity and ripening stage from photographs or videos, with the ability to detect them even when they still have a green coloration.
The same approach can be applied during the winter period to estimate the density of spurs or buds on trees, generating diagnoses on the number of expected productive structures in a plot. This is particularly useful for assessing losses associated with climatic events, such as heavy rainfall, or with management practices such as thinning. By mapping this information in different areas of the plot, it is possible to identify spatial patterns and production behaviors that would otherwise go unnoticed, enabling the early detection of anomalous areas within the farm.
Image and video sequences make it possible to estimate tree morphology over time. However, the presence of occlusions, the dynamic nature of foliage and uncontrolled lighting variations introduce ambiguities that require more robust reconstruction methodologies. In recent years, AI models have emerged that are capable of generating three-dimensional representations directly from images, learning from the information contained in pixels and from variations between multiple viewpoints. This makes it possible to obtain more coherent reconstructions that are suitable for quantitative measurements, even under conditions of occlusion and changes in lighting.
A complete reconstruction also makes it possible to extract specific components of the scene in an assisted way through segmentation techniques. This allows structures of interest to be isolated in order to estimate their spatial distribution and support the assessment of variables associated with productivity and quality. Field data collection is carried out using two camera systems with complementary objectives: one focused on the three-dimensional reconstruction of the complete tree, to model its architecture and the spatial distribution of fruit, and another, with a stereo camera, focused on the detailed reconstruction of individual cherries. This combination of scales makes it possible to integrate the structural analysis of the tree with the fine geometric characterization of fruit caliber and shape.
From a scientific perspective, these reconstructions allow morphological analyses of the tree to be carried out over time. By capturing images at different phenological stages, it is possible to “freeze” specific moments in tree development and assess its structural evolution between seasons or during the annual production cycle.
A reconstructed cherry therefore corresponds to the digital three-dimensional representation of the fruit obtained through the processing of multiple images acquired from different angles. This 3D model makes it possible to visualize and quantify geometric attributes such as shape, volume and peduncle structure, as well as to analyze any asymmetries and improve caliber estimation.
The work developed at UOH goes beyond the simple counting of fruit. Based on three-dimensional reconstructions of trees and fruit, together with the use of hyperspectral images, digital models are generated that make it possible to quantify not only the number of fruit, but also their growth and ripening level. In this way, the data collected in the field are transformed into structured information that feeds growth and quality models, facilitating better-founded agronomic decisions.
Early yield estimation is essential for planning logistics, labor and commercial commitments. The 3D computer vision-based approach makes it possible to reconstruct branches and orchard sectors, estimating the volume, distribution and evolution of fruit caliber. By integrating this information with agroclimatic variables, more robust projections of production potential are generated. It is not just about counting fruit, but about understanding its growth dynamics.
Crop load regulation largely defines the balance between quantity and quality: excessive load compromises caliber and uniformity, while insufficient load penalizes yield. The models developed make it possible to assess the real growth status of the fruit in different positions on the tree and at different times in the cycle, supporting regulation decisions with a stronger technical foundation. In parallel, through hyperspectral images integrated into 3D models, the project aims to correlate spectral information with internal variables, such as firmness and Brix degrees, in order to estimate ripening non-destructively, directly on the tree. Having this information would make it possible to adjust harvest windows, optimize operations and reduce the risk that fruit fails to meet commercial standards in destination markets. In a scenario of climate change and high commercial demands, these tools are strategic for the competitiveness of the export industry.
In sweet cherry orchards, every square meter can present different environmental conditions: areas with greater shading, sectors exposed to wind or areas vulnerable to frost due to proximity to roadside edges. These variations generate significant impacts on the phenological development of the crop, affecting fruit quality, caliber and uniformity. Having reliable real-time data is essential to make visible what previously remained invisible: the microclimates that determine the yield of each plant.
The enabling technology used in this project is LoRaWAN, a long-range, low-power wireless communication protocol belonging to the LPWA (Low Power Wide Area) family. Its wide coverage, up to 15 km, and the high energy autonomy of the nodes, up to 10 years, make it possible to maintain extensive networks with minimal infrastructure.
Thanks to these characteristics and regional funding (FIC IDI 40059070-0), sensors for temperature, humidity and soil-related variables have been installed in different sectors of each participating farm. The devices transmit real-time data to a central antenna, from which the information is sent to the database located at UOH, where AI models interpret it and generate intra-farm climate maps. The network currently operates in the municipalities of San Fernando, Rengo, Requínoa, Graneros, Peumo and Las Cabras, making it possible not only to compare conditions between farms, but also to identify thermal contrasts between adjacent sectors within the same field, down to the level of the individual tree.
During the latest season, this network made it possible to accurately determine the levels of chill accumulation in 24-hour intervals and to verify that this accumulation was not homogeneous between sectors. This evidence enabled the differentiated adaptation of management strategies, such as the localized application of cyanamide or bud-break stimulating products in the most delayed areas, aligning agronomic decisions with the local thermal reality.
With the aim of ensuring that these technologies have a real impact on production management, a digital platform has been developed that allows the farmers involved in the project to access periodic reports on the environmental conditions of their crops. The reports include key indicators for agronomic decision-making: accumulation of chill hours (CH), degree-days (DD), average temperature, environmental humidity and variables associated with soil status. This systematic availability makes it possible to more accurately assess phenological development and anticipate the risks associated with climatic events.
In a complementary way, AI-based methodologies for analyzing structures in sweet cherry trees will be integrated into a web application aimed at growers, allowing them to visualize statistics, georeferenced data and graphic representations of the production status of orchards. In this way, farmers will be able to generate reports by integrating information from both climate monitoring and image analysis, creating a continuous diagnostic tool to support decisions in crop management.
The implementation of these technologies in real production contexts still faces major challenges. The quality and volume of data collected in the field are critical aspects, as computer vision models require large, heterogeneous and correctly labeled datasets. Building them is a labor-intensive process that requires constant coordination with growers. In addition, one of the main challenges is scalability: the methodologies developed on pilot farms must be transferred to production systems with different varietal configurations, densities and surface areas, which requires replicable acquisition protocols and sufficiently generalizable models. For medium and small growers, access to these technologies will also depend on the development of scalable and cooperative service models capable of reducing the entry barrier.
In the short and medium term, the most significant evolution points to the progressive integration of data flows of different kinds, phenological, climatic, spectral and structural, into unified platforms capable of operating not only as diagnostic tools, but also as systems to support strategic planning for the season. In this scenario, the role of the grower and the agronomic consultant will not be replaced, but transformed: experiential management knowledge will remain indispensable, but it will have to be combined with the ability to interpret quantitative data and interact with decision-support platforms. At the production model level, the continuous adoption of precision technologies has the potential to redefine the sector’s competitiveness standards, enabling traceability from tree to shipment and quality certification, with quantitative support, in increasingly demanding destination markets.
Rodrigo Versachae
Academic at USM, scientific director of the FIC Cherry Project. Karen Mesa, academic at the ICA3 Institute of Agronomic and Veterinary Sciences. Jaime Varas, agronomist from PUC, executive coordinator of the FIC Cherry Project, ICI-UOH
Image source: Stefano Lugli
26 Aug 2025
All methods to determine the optimal harvest maturity of sweet cherries: color, weight, soluble solids, acidity, and taste. Combined physical and chemical approach with data from Turkey, Afghanistan, USA, and Denmark to maximize fruit quality and market value.
22 Apr 2025
A Romanian study analyzes the correlation between fruit quality and cracking in three sweet cherry cultivars: Special, Tentant, and Severin. The results highlight significant differences in weight, pH, color, °Brix, and cracking index, with implications for varietal selection.
09 Jun 2026
Tart cherry juice clarification with PES membranes and pectinase-based enzymatic pretreatment improves clarity and stability, limits fouling and protects polyphenols, anthocyanins and antioxidants, offering useful guidance for more sustainable and efficient processing methods.
09 Jun 2026
In Chile’s cherry orchards, the University of O’Higgins combines AI, LoRaWAN sensors, computer vision, hyperspectral imaging and 3D models to estimate yield, fruit size, ripening and microclimates, improving quality, harvest planning and agronomic decisions.
