Method uses artificial intelligence to separate coffees before roasting

According to an article published in the scientific journal Computer and Electronics in Agriculture🇧🇷 scientists University of Sao Paulo (USP) managed to develop method with artificial intelligence (AI) to select grains Coffee special and traditional🇧🇷 Brazilian researchers from the Center for Nuclear Energy in Agriculture (Cena) at USP and the Luiz de Queiroz School of Agriculture (Esalq).

New engine uses multispectral imaging technique (MSI)based on reflectivity and autofluorescence, and machine learning to separate grains in the production process — the process is real time and does not require roasting. They use artificial intelligence to analyze a series of images of the same region at different wavelengths, thus succeeding in classifying the grains.

Adoption of the new technique for classifying coffee beans is still very recent in the coffee industry, according to Winston Pinheiro Claro Gomes, first author of the study and PhD student at Cena. However, it is often used for nitrogen mapping in fields and for detecting necrosis, pests and diseases that may affect coffee during cultivation.

smart grain separation

“First, monochrome reflection images were captured. [reflectância]🇧🇷 Next, autofluorescent images were acquired. Therefore, we extracted information referring to the interests in the software of the equipment used to build the mathematical models that classify the samples and provide us with the result (from Fundação de Amparo to São Paulo State Research).

Typically, the process of separating specialty coffee beans requires three checks: a raw coffee sample; another example of roasted coffee; and a sensory test by taste. According to the Specialty Coffee Association of America (SCAA), coffee quality ranges from 0 to 100 points, but the new method still cannot fully detail the grain score.