The process of selecting specialty coffee involves three types of inspection. Two are physical and include samples of both raw and roasted coffee. The third is sensory and involves tasting the drink. Certification is by the Specialty Coffee Association of America (SCAA).
According to SCAA guidelines, coffee quality is measured on a decimal scale from zero to 100. A specialty coffee must achieve 80 or more points. The grower sends a sample of raw beans to three cuppers (tasters) who roast and prepare coffee from each batch, again in accordance with SCAA standards, before providing a report.
Brazilian scientists from the Center for Nuclear Energy in Agriculture (CENA-USP) of the University of São Paulo, in collaboration with colleagues from the Luiz de Queiroz College of Agriculture (ESALQ-USP) and the Computing Center of the Federal University of Pernambuco (UFPE), have developed a procedure for the selection of developed coffee beans based on multispectral imaging and machine learning. The process does not require roasting and can be performed in real time during the production process. It avoids possible human error, although it relies on expensive equipment. An article about the new method recently appeared in Computers and Electronics in Agriculture.
“Specialty coffees are often harvested selectively, meaning only the ripe red cherries are picked. They are harvested one by one by hand. If a specialty coffee grower is harvesting green beans or using manual and/or mechanized strip picking at any point, this process can result in a standard commercial harvest,” said Winston Pinheiro Claro Gomes, first author of the article. Gomes is a PhD student in chemistry at CENA-USP, with Wanessa Melchert Mattos and Clíssia Barboza da Silva as supervisors.
“In our method, we separate beans that are considered a specialty and a commercial standard using a combination of multispectral imaging and mathematical algorithms that process the data provided by the images,” Gomes explained. “Specialty coffees have to score between 80 and 100, but our model can’t tell whether beans score 80 or 90. That would require machine learning with sampling for each score to specify these categories in the mathematical model.”
The research was conducted with support from a Young Investigator Grant awarded to Barboza da Silva, the penultimate author of the article, and a Regular Research Grant awarded to Mattos, the article’s last author.
Multispectral methodology
The team used a multispectral imaging (MSI) technique based on reflection and autofluorescence, capturing images of the same object at different wavelengths, followed by a machine learning model to classify beans using information gleaned from the images.
“The use of MSI in the coffee industry is very new. It is mainly used to map nitrogen in coffee groves, detect necrosis in beans, and detect pests and diseases in plants, according to the literature on the subject,” Gomes said.
Researchers analyzed 16 samples of specialty and standard crop green beans grown in the states of Minas Gerais and São Paulo. Ten of the specialty coffee beans (Coffea arabica) were from the 2016/17 crop grown in the Alta Mogiana region. They were judged at the Alta Mogiana Coffee Contest 2017 and supplied by the region’s Association of Specialty Coffee Producers. The other six samples were taken from standard commercial cultures purchased in bulk from the local market.
For each sample, 64 beans were randomly separated without prior treatment, yielding a total of 1,024 beans (384 standard beans, 640 special beans), and used for machine learning calibration, validation, and testing.
Gomes summarized the procedure as follows: “We put the beans in a Petri dish and put them in the device, which is a sphere containing LEDs, optical filters and a camera. The camera descended over the samples until they were completely covered and took pictures after homogeneous and diffuse illumination at different wavelengths. Monochrome reflectance images were taken first, followed by autofluorescence images, after which information about the regions of interest was extracted by the onboard software and used to create the algorithms that classified the samples and gave us the results.”
A principal component analysis (PCA) was then performed to examine the variables influencing the differences between specialty and standard coffees. Researchers ran four machine learning algorithms, with the Support Vector Machine (SVM) proving to be the best, and used to calculate coefficients to estimate the key variables.
fluorescence
Specialty beans were more uniform in shape in the visible spectrum (RGB) images, while standard beans were more intense in the autofluorescence images. “Our mathematical model and algorithms use signal intensity information from fluorescence images. It can happen that a compound present in beans becomes more excited at a certain wavelength. A more or less intense fluorescence signal can, for example, also relate to a change in the concentration of a compound in beans. The model we chose was the one that performed best at distinguishing between specialty and standard coffee beans. In this model, the most important information for constructing separation boundaries came from the green fluorescence. We therefore decided to analyze the individual compounds that are inherently green fluorescent and attempted to assign some fluorescent compounds that might affect the coffee classification separation process,” said Gomes.
Green fluorescence, a biological marker represented by green light in the visible spectrum, was analyzed for 10 phenolic compounds and the data showed that catechin, caffeine and certain acids (4-hydroxybenzoic acid, sinapic acid and chlorogenic acid) reacted intensely upon excitation with blue Light at 405 nanometers (nm) emitting energy at 500 nm. This autofluorescence data (excitation/emission at 405/500 nm) contributed most to distinguishing specialty green beans from standard green beans.
“These are chemical species associated with aromatic groups that absorb energy with respect to a specific wavelength. With autofluorescence-based methods, differences in the concentrations of these chemical species in specialty and standard coffees can be used to distinguish between the two groups,” Gomes said.
Differences in the levels of these compounds are typically used to distinguish between specialty and standard coffee beans. “For my master’s thesis, I examined the chemical composition of these samples, and while there were no differences in the chemical species, we found differences in their concentrations, particularly the concentrations of chlorogenic acid and caffeine,” he said.
The next steps, according to Gomes, will be to obtain samples from each of the SCAA-defined specialty coffee rating levels (not an easy task) and classify the beans according to their rating. “In Brazil, coffees are rated 90-92 at most. Higher is hard to find. Only imported coffee, for example from Ethiopia, achieves 100 points. In my PhD, I am trying to classify beans based on X-ray images and have decided to increase the number of samples and the range of analysis by including imported beans,” he said.
About the Sao Paulo Research Foundation (FAPESP)
The São Paulo Research Foundation (FAPESP) is a public body with a mission to support scientific research in all fields of knowledge by awarding grants, fellowships and grants to researchers affiliated with universities and research institutions in the state of São Paulo, Brazil . FAPESP recognizes that the very best research can only be achieved by collaborating with the best researchers at the international level. Therefore, it has established partnerships with funding agencies, universities, private companies and research institutes in other countries known for the quality of their research and has encouraged scientists funded by its fellowships to further develop their international collaborations. You can learn more about FAPESP at www.FAPESP.br/en and visit the FAPESP news agency at www.agencia.FAPESP.br/en to keep up to date with the latest scientific breakthroughs that FAPESP is making through its many programs, Awards and research centers supported. You can also subscribe to the FAPESP news agency at https://agencia.FAPESP.br/subscribe
/Public release. This material from the original organization/author(s) may be post-date in nature and may be edited for clarity, style and length. The views and opinions expressed are those of the author(s).