Smart Tech for Sizing Fruit and Detecting Tree Thirst

Jim Crocker
14th July, 2025

Smart Tech for Sizing Fruit and Detecting Tree Thirst

To enable the digital fruit size estimation method, plastic reference objects were adhered to Satsuma mandarin (Citrus unshiu) fruits (a) to serve as a fixed scale for the fruit sets (b) during image acquisition.

Image adapted from: Komiya et al. / CC BY (Source)

Key Findings

  • Researchers in Japan developed a smartphone-based AI system to measure Satsuma mandarin fruit size, which helps determine how much water the trees need
  • This system accurately estimates fruit diameter and reliably links fruit growth to the tree's water stress, outperforming traditional manual measurements
  • The real-time fruit growth data from this method can be used to automate irrigation, optimizing water use and improving the quality of high-sugar mandarins
Effective water management is critical for modern agriculture, particularly for high-value crops like Satsuma mandarins. Precise control over irrigation, especially during key growth periods like summer and fall, directly influences fruit quality, such as sugar content, and overall yield. Traditionally, assessing a tree's water stress has been a labor-intensive process, often relying on manual observations or less precise methods. The increasing demand for automated irrigation systems highlights the need for digital, real-time data on plant health and water status. Addressing this challenge, recent research from the National Agriculture and Food Research Organization[1] has developed a novel smartphone-based method for estimating fruit size, which can then be used to infer tree water stress. The core idea behind this approach is that fruit growth is closely linked to the amount of water a tree receives. By accurately tracking how fruit develops, growers can gain insights into the tree's water needs. The method developed by the National Agriculture and Food Research Organization leverages advanced image processing techniques. It combines Mask RCNN, a type of artificial intelligence that can not only identify objects in an image but also precisely outline their boundaries, with a regression model. Mask RCNN is used to accurately detect and segment individual fruits from the background in images captured by a smartphone camera. Once the fruit is isolated, its dimensions can be measured. A regression model is then employed, which is a statistical tool used to establish a mathematical relationship between different variables. In this case, it helps to accurately estimate the fruit's actual diameter from its image and, crucially, to correlate fruit growth and measured diameter with the tree's water stress levels. The study reported impressive accuracy, with the estimated fruit diameters closely matching actual measurements, and strong correlations found between fruit growth, measured diameter, and water stress. The concept of using machine vision for in-field fruit sizing has been explored in various contexts. For instance, earlier work investigated methods for sizing mangoes in the field using technologies like RGB-D (depth) cameras, stereo vision, and Time of Flight (ToF) laser rangefinders[2]. That research highlighted the practicality of machine vision for estimating fruit size, though it noted limitations such as poor performance of RGB-D cameras in direct sunlight. It involved complex calibration, fruit detection using techniques like histogram of oriented gradients (HOG) and Otsu's method, and calculations based on depth information and the thin lens formula. While effective, such systems often required specialized equipment. Building on the accessibility of mobile phones, another study developed an Android application called "FruitSize"[3], which allowed farm staff to measure fruit size in the field using a phone camera. This application relied on imaging fruit against a backboard with a scale and used image processing techniques from the OpenCV library to segment fruit and calculate sizes, achieving good accuracy for various fruits including mandarins and mangoes. This demonstrated the feasibility of using ubiquitous smartphone technology for practical agricultural measurements. The current study from the National Agriculture and Food Research Organization takes this smartphone-based approach a significant step further. While previous efforts focused on fruit sizing for purposes like yield estimation or harvest planning[2][3], this new research directly links the estimated fruit size to a critical physiological parameter: tree water stress. By integrating the advanced Mask RCNN for precise fruit detection and a robust regression model, the system provides a highly accurate way to monitor fruit development. This real-time data on fruit growth, acting as an indicator of water availability, can then be fed directly into automated irrigation systems. This allows for dynamic adjustment of water delivery based on the actual needs of the trees, optimizing water usage, reducing waste, and ultimately leading to the production of higher quality fruits, such as the desired high-sugar Satsuma mandarins. This represents a practical and cost-effective solution for precision agriculture, leveraging readily available smartphone technology to make informed, real-time decisions about water management.

FruitsAgriculturePlant Science

References

Main Study

1) Development of a fruit size estimation method using Mask RCNN for water stress estimation in Satsuma mandarin trees

Published 11th July, 2025

https://doi.org/10.1371/journal.pone.0324278

Related Studies

2) On-Tree Mango Fruit Size Estimation Using RGB-D Images.

https://doi.org/10.3390/s17122738

3) In Field Fruit Sizing Using A Smart Phone Application.

https://doi.org/10.3390/s18103331


Related Articles

An unhandled error has occurred. Reload 🗙