In Brazil, researchers at the Agricultural Nuclear Energy Center (CENA) and the Luis Deceiras Agricultural University (ESALQ), which are part of the University of São Paulo (USP), have developed a methodology based on automated artificial intelligence. Streamline seed quality analysis. This is required by law and is currently a manual process performed by certified analysts of the Ministry of Agriculture.
Light-based technology consisted of chlorophyll fluorescence and multispectral imaging. From plants related to both crops and experimental models, researchers chose tomatoes and carrots that were produced in different countries and seasons and subjected to different storage conditions. They used the seeds of gaucho and Tina commercial tomato varieties produced in Brazil and the United States, and the seeds of Brasilia and Francine carrot varieties produced in Brazil, Italy and Chile.
This choice was based on the economic importance of these edible crops, which are in high demand worldwide, and the difficulties growers face in collecting seeds. In both tomatoes and carrots, the maturation process is not uniform because the plants bloom continuously and seed production is asynchronous. Therefore, seed lots may contain a mixture of immature seeds and mature seeds. The presence of immature seeds is not easily detected by visual methods, and machine vision-based technology can minimize this problem.
Researchers compared the results of non-destructive analysis with the results of traditional germination and vitality tests, which are destructive, time-consuming and labor-intensive. In the germination test, a seed analyst separates the sample, sows it to germinate under appropriate temperature, water and oxygen conditions, and confirms the final amount of normal seedlings produced according to the rules set by the Ministry of Agriculture. The vitality test is complementary and more sophisticated. Most commonly, it is based on seed response to stress and seedling growth parameters.
In addition to the difficulties mentioned, traditional methods are time consuming. For tomatoes and carrots, for example, it can take up to two weeks to get results, but this is also largely subjective, depending on analysts’ interpretation. “Our suggestion is to use chlorophyll fluorescence and multispectral imaging to automate the process as much as possible and analyze seed quality, which avoids all the usual bottlenecks,” CENA said. -Articles on research published in USP Researcher Frontier of plant science..
Silva is the Principal Investigator of a project supported by the São Paulo Research Foundation-FAPESP. The lead author of this article is Patricia Galletti, who conducted her research as part of her master’s research and won the Best Poster Award at the 7th American Seed Congress in 2019. There she announced the partial results of the project.
Chlorophyll as a quality marker
Chlorophyll is present in seeds and provides the energy to store the nutrients (lipids, proteins, carbohydrates) needed for development. Chlorophyll breaks down when it performs this function. “But if the seed does not complete the maturation process, this chlorophyll remains in it. The less residual chlorophyll, the more advanced the maturation process and the higher the quality of the nutrients contained in the seed. If the seed is high in chlorophyll, the seed Is immature and its quality is poor, “Silva said.
When light of a particular wavelength hits the seed chlorophyll, this energy is not transferred to another molecule and instead re-emits light of another wavelength. That is, it fluoresces. She explained that this fluorescence can be measured. You can use red light to excite chlorophyll and use a device that converts it into an electrical signal to capture fluorescence and produce images containing gray, black, and white pixels. The bright areas correspond to high levels of chlorophyll, indicating that the seeds are immature and unlikely to germinate.
In multispectral imaging, LEDs (light emitting diodes) emit light in the visible part of the spectrum and invisible light (UV and near infrared). To analyze seed quality based on reflectance, researchers used 19 wavelengths and compared the results with quality evaluation data obtained by conventional methods. Best results were obtained using near-infrared rays for carrot seeds and UV for tomato seeds.
Seeds contain proteins, lipids, and sugars that absorb some of the light emitted by the LEDs and reflect the rest. The reflected light is captured by a multispectral camera, and the captured image is processed to separate the seed from support for devices that support zero-valued black pixels, while the seed is grayscale. The pixel values in the seed image correspond to its chemical composition.
“We don’t get average results for our samples. We perform a separate extraction for each seed,” says Silva. “The higher the amount of specific nutrients in a seed, the more light of a particular wavelength is absorbed and less reflection. Seeds with lower nutrient content have fewer light-absorbing molecules. This means higher reflectance, which depends on the components that behave differently depending on the wavelength of light used. “
The algorithm identifies the wavelength that gives the best results. This process provides information about the chemical composition of the seed, from which its quality can be inferred.
For researchers, this was not enough to reach the imaging stage, as it is still an operation that requires human observation. “Next, we developed a series of statistical and mathematical methods, chemometrics, used to chemically classify materials,” says Silva. “The idea was that equipment needed to classify quality based on captured images.” The method used by scientists in this study is widely used in the medical and food industries.
We then leveraged machine learning to test models created using chemometrics. “We taught the model to distinguish between high-quality and low-quality seeds. We used 70% of the data for training the model and the remaining 30% for validation,” says Silva. Quality classification accuracy ranged from 86% to 95% for tomato seeds and 88% to 97% for carrot seeds.
Given the speed of image capture, the two main techniques were accurate and time-saving. The chlorophyll phosphor captured one image per second while the multispectral imaging analyzer processed 19 images in 5 seconds.
The unexpected results produced during the course of the project proved to be very important. Chlorophyll fluorescence and multispectral imaging are also efficient methods of plant variety screening, which are an important part of seed lot evaluation to avoid economic losses. “Growers buy seeds with the expectation of a certain yield, but if seeds with different genetic characteristics are not properly isolated, production will be affected,” says Silva.
Screening is currently being conducted by trained analysts with the skills required to grade seeds by color, shape, size, and, if possible, molecular markers. This study proved that both methods were efficient in separating carrot varieties, but for tomato varieties, multispectral imaging was inadequate.
“This study has produced new results regarding the use of fluorescence to screen varieties,” Silva said. “No previous studies have been found in which fluorescence was used for this purpose. Some studies have shown that multispectral imaging is efficient for this purpose, but we use it. That’s not the case with the equipment that was used. “
A good way to transfer the knowledge generated by the research to the production sector is to get the company to develop equipment for sale to seed producers, Silva said. “For example, it is possible to use our findings to characterize the quality of tomato seeds using only UV light and develop equipment to market,” she speculated.
Studies point to challenges in harvesting weed seed management in the Pacific Northwest
For more information:
Patrícia A. Gallettietal, integration of optical imaging tools for rapid and non-invasive characterization of seed quality: tomatoes (Solanum lycopersicum L.) and carrots (Daucus carota L.) as study cases, Frontier of plant science (2020). DOI: 10.3389 / fpls.2020.577851
Quote: Crop seed analysis obtained on March 19, 2021 from https: //phys.org/news/2021-03-technique-based-artificial-intelligence-automation.html by technology based on artificial intelligence (2021, March 19th) will be automated
This document is subject to copyright. No part may be reproduced without written permission, except for fair transactions for personal investigation or research purposes. The content is provided for informational purposes only.
Artificial intelligence-based technology enables automation of crop seed analysis
Source link Artificial intelligence-based technology enables automation of crop seed analysis