In Brazil, researchers affiliated with the Center for Nuclear Power in Agriculture (CENA) and the Luiz de Queiroz Agricultural College (ESALQ), both of which are part of the University of São Paulo (USP), have developed a development based on artificial intelligence to streamline seed quality automation and monitoring, a process that is currently required by law and carried out by hand by analysts certified by the Ministry of Agriculture.
The group used light-based technology such as those used in plant and cosmetics analysis to obtain images of the seeds. They then turned to machine learning to automate the process of image interpretation, reducing some of the difficulties of conventional methods. For example, for many species, optical imaging technology can be applied to a whole batch of seeds rather than just samples, as is currently the case. Moreover, the method is non-invasive and does not destroy the observed results or generate residues.
Light-based techniques included chlorophyll fluorescence and multivariate imaging. Among relevant plants such as both crops and experimental models, the researchers selected tomatoes and carrots produced in different countries and seasons and applied to different storage conditions. They used seeds of the commercial tomato varieties Gaucho and Tyna extracted in Brazil and the USA, and seeds of the carrot varieties Brasilia and Francine extracted in Brazil, Italy and the United States. -Sile.
The choice was based on the economic importance of these food crops, for which world demand is high and rising, and on the difficulties faced by growers in collecting their seeds. In both tomato and carrot, the ripening process is not uniform as the plants are continuously flowering and seed yield is asynchronous, so that seed wounds have a mixture of abnormal and mature seeds. Immediate seed presence is not easy to detect by visual methods, and methods based on machine vision can alleviate this problem.
The researchers compared the results of their non-destructive analysis with the results of traditional germination and vigor tests, which are destructive, time-consuming and labor-intensive. In the germination test, seed analyzers separate samples, send them to germination in favorable conditions, water, and oxygen, and determine the amount of normal seed to be extracted according to rules of the Ministry of Agriculture. Live tests are complementary and more sophisticated. The most common are based on the seed’s response to weight gain and germination parameters.
In addition to the problems mentioned, traditional methods take time. As for tomatoes and carrots, for example, it can take up to two weeks to get results, which is also largely personal, according to the analyst’s explanation. “Our proposal is to automate the process as much as possible by using chlorophyll fluorescence and multivariate imaging to monitor seed quality. This will avoid the normal bottles,” said Clíssia Barboza da Silva, researcher at CENA-USP and one of the authors of an article on the study published in Boundaries in Plant Science.
Silva is the lead researcher for the project with support from the São Paulo Research Foundation – FAPESP. The lead author of the article is Patrícia Galletti, who conducted the study as part of her master’s study and won the Best Poster Award in 2019 at the 7th American Seed Congress, where she presented part of the project’s results.
Chlorophyll as an indicator of quality
Chlorophyll is present in seeds, where it provides energy for storing nutrients needed for development (lipids, proteins, and carbohydrates). Once it has accomplished this function, the chlorophyll breaks down. “However, if the seed does not complete the maturation process, this chlorophyll will remain inside. The less chlorophyll is left, the more advanced the maturation process will be and the larger and higher the quality. if there is a lot of chlorophyll., the seed is abnormal and its quality is poor, “said Silva.
If light from a particular wave shines on the chlorophyll in seeds, it does not transfer this energy to another molecule but instead redistributes the light at another wave, meaning it is flowering. This fluorescence can be measured, she explained. Red light can be used to stimulate chlorophyll and capture the fluorescence using a device that converts it to an electrical signal, producing an image consisting of gray, black and white pixels. The lighter areas respond to higher levels of chlorophyll, indicating that the seed is abnormal and unlikely to grow.
In multi-view images, LEDs (light-emitting codes) emit light in the visible part of the spectrum as well as invisible light (UV and near-infrared). To study the quality of shadow-based seeds, the researchers used 19 waves and compared the results with quality assessment data obtained through traditional methods. The best results were obtained by using near infrared in the case of carrot seeds and UV in the case of 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 multi-view camera, and the captured image is processed to separate the seeds from the support in the device, which corresponds to black pixels with a value of zero, while as the seeds are gray. The pixel values in a seed image correspond to its chemical content.
“We don’t work with an average yield for a sample. We do an individual extraction for each seed,” Silva said. “The greater the amount of a particular nutrient in a seed, the more light it absorbs of a particular wave so that there is less reflection. There are fewer molecules that are contains light in seeds with lower nutrient content, which means that its shade is higher, although this varies according to its components, which behave differently. varies depending on the wave of light used. “
An algorithm identifies the wavelength that gets the best result. The process provides information about the chemical condition of the seed, from which its quality can be determined.
For the researchers, it was not enough to reach the level of images, as this is still an activity that requires human observation. “We then applied chemometrics, a set of statistical and mathematical methods used to chemically classify materials,” Silva said. “The idea was that the equipment should be quality classified according to the image it captured.” The techniques used by the scientists in this study are widely used in medicine and the food industry. .
Next, they influenced the learning of machines to test the models created using chemometrics. “We learned the model to identify high-quality and low-quality seeds. We used 70% of our data to train the model, and we used the remaining 30% for validation,” Silva said. Quality classification accuracy ranged from 86% to 95% for tomato seeds, and from 88% to 97% for carrot seeds.
The two main methods were both accurate and time-saving, due to the speed of image capture. The chlorophyll fluorescence instrument captured one image per second, and the analyzer processed multiple images of 19 images in five seconds.
An unanticipated outcome achieved in the course of the project was particularly significant. Chlorophyll fluorescence and multivariate imaging are also effective methods for plant mixture screening, an essential part of seed injury assessment to avoid economic loss. “Growers buy seeds with the expectation of a particular crop yield, but productivity will be affected if seeds with different genetic traits are not properly segregated,” Silva said.
Screening is currently performed by analysts trained in the skills needed to rank seeds by color, shape, and size, as well as molecular markers where possible. In the study, both methods were effective for separating carrot varieties but multivariate images were unsatisfactory for tomato varieties.
“The study yielded novel results regarding the use of fluorescence to screen types,” Silva said. “We did not find any prior study in which fluorescence was used for this purpose. Some studies show that multivariate imaging is effective for this purpose, but not with the instrument we used. . “
A good way to transfer the knowledge gained from the research to the production sector, Silva said, would be for companies to develop the equipment for sale to seed producers. “It would be possible to use the results of our research to develop an instrument that used only UV light to identify and bring to market the quality of tomato seeds, for example,” she said.
A study highlights the challenges of harvesting herb seeds at harvest time in the Pacific Northwest
Patrícia A. Galletti et al, Integrating optical imaging devices for fast and non-invasive character of seed quality: tomato (Solanum lycopersicum L.) and Carrot (Daucus carota L.) as Case Studies, Boundaries in Plant Science (2020). DOI: 10.3389 / fpls.2020.577851
Citation: Artificial intelligence-based innovation allows automation of crop seed self-analysis (2021, March 19) on March 19, 2021 retrieved from https://phys.org/news/2021-03-technique- based-artificial-intelligence-automation.html
This document is subject to copyright. Other than any fair treatment for the purpose of scrutiny or private investigation, no part may be reproduced without written permission. The content is provided for informational purposes only.