According to an article published in the scientific journal Computers and Electronics in Agriculture scientists of University of Sao Paulo (USP) managed to develop a method with artificial intelligence (AI) to select grains of coffee special and traditional The Brazilian researchers are from the Center for Nuclear Energy in Agriculture (Cena) and the Luiz de Queiroz Higher School of Agriculture (Esalq), at USP.
The new engine uses the multispectral imaging (MSI) technique based on reflectance and autofluorescence, and machine learning to separate the grains during the production process — the process is in real time and does not require roasting. They use artificial intelligence to analyze a set of images of the same region at different wavelengths and thus manage to classify the grains.
According to the first author of the study and a doctoral candidate at Cena, Winston Pinheiro Claro Gomes, the adoption of the new technique for classifying grains is still very recent in the coffee industry. However, it is often used in nitrogen mapping in plantations and in detection of necrosis, pests and diseases that can affect coffee during cultivation.
Source: Unsplash
Intelligent grain separation
“First, monochromatic reflection images were captured [reflectância] Then, autofluorescence images were captured. Then, we extracted information regarding the regions of interest from the equipment’s software, which were used to build the mathematical models that classified the samples and provided us with the result,” Gomes told the Pesquisa para Inovação website, linked to FAPESP (Fundação de Amparo to the Research of the State of São Paulo).
Normally, the process of separating beans from specialty coffees requires three checks: a sample of raw coffee; another sample of roasted coffee; and a sensory test through tasting. According to the Specialty Coffee Association of America (SCAA), the quality of coffee ranges from 0 to 100 points, however, the new method still cannot accurately detail the score of the beans.