Sap flows in the crop are an important indicator of plant development. Students at Inholland University of Applied Sciences worked in the Ideal Research Greenhouse Lab to measure and visualize the sap flows in a cherry tomato crop, whereby Machine Deep Learning has been used to predict the growth of the head thickness of the plants based on these data, they wrote in a research report.
(Archive) photos: Inholland
Within the project, which was monitored using heat-balance sensor technology, algorithms were developed that monitor the moisture balance of the plant.
The sap flows (Xylem) and their balance are in practice among the most difficult to measure and to control parameters, the researchers said in the report. "Plants can be grown optimally based on proactive response to evaporation of moisture from the plant."
A model has been developed to control, automate, regulate and optimize the development process of the plant while wastage of raw materials such as water, nutrients, etc. is minimized or even prevented.
Sap flow data
Recent developments in Machine Learning (ML) and especially so-called Deep Learning (DL), gave the researchers, in their own words, 'powerful new analytical tools'. "With these techniques it is possible to use sap flow data and stem thickness to predict and adjust plant development."
According to the researchers, the major advantage of Machine Learning (ML) techniques is that they are able to independently solve complex problems by using various data sets and sources. "This allows for better decisions and action in contemporary scenarios with no or minimal human intervention."
Data vs. practice
A comparative study was conducted in which various ML methods were used to analyze plant data and benchmark them for effectiveness and accuracy in relation to the real situation. A dashboard has been developed to give a clear overview of the predictions that the models provide and to enable the grower to influence the development of his crops. It allows the grower to see whether his crop is developing in a healthy way or whether there are pathogens in the crop.
The research took place within the framework of the ERDF Evergreen Project in the Ideal Lab Greenhouse in the World Horti Center in Naaldwijk. Cherry tomatoes had been planted there. In the greenhouse, sap flows and stem thickness sensors from 2GROW were placed on the stem of the tomato plant. This yielded data that could be used.
To be precise, it concerned PhytoClip sensors (introduced at the end of 2019). The new sensor made it possible to measure smaller plants, while larger sensors can be applied for larger crops. By combining the PhytoClip sensor with the PhytoStem sensors it becomes possible to 'measure almost any type of plant'.
Proposal for alternative contactless sensor
Until now, the only sensors on the market are sensors that make physical contact with the plant. However, establishing physical contact causes stress to the plant. This stress may cause a plant to stop its growth for two days, so that it cannot be used as a reference for a population of plants and furthermore will also become a less strong plant.
Students at Inholland have therefore conducted research into contactless alternatives and came up with a concept based on an optical light lock that can be moved along the stem of the plant. In this way, the thickness of the stem can be measured without contact but with sufficient precision. However, further elaboration is necessary to arrive at a prototype.
View the research report here (in Dutch), for which Ir. Amora Amir, Dr. Marya Butt and Ir. Cees Jeroen Bes from Inholland were the project supervisors.