Brazilian researchers used AI and multi-spectral imaging technology to sort coffee beans into two categories: specialty and standard. Photo courtesy of Winston Pinheiro Claro Gomes.
Sorting out specialty coffee beans takes time and human effort, but a new approach could enable real-time sorting during the production process. Brazilian researchers published a study on applying multispectral imaging (MSI) and machine learning to classify green coffee beans as specialty or traditional commercial.
Specialty coffee is typically defined according to technical specifications outlined by the Specialty Coffee Association. A coffee must score at least 80 points on the SCA's 100-point scale to earn the specialty designation. Cuppers trained to use SCA standards inspect, roast and taste samples of coffee to generate a quality score. Using technology for this process would mean bean samples do not need to be sent for testing or roasted.
The study, published in Computers and Electronics in Agriculture, describes how researchers used MSI and machine learning, a subtype of artificial intelligence, to identify the difference between specialty and traditional coffee beans. Using MSI, researchers took images of the coffee beans at different wavelengths of light.
Four different machine-learning algorithms analyzed the images. The researchers found that microscopic molecules known as fluorophores, including caffeine, catechin, and chlorogenic acids among others, are key to differentiating between traditional and specialty beans. The process is non-destructive and takes place in real-time.
While this research does show promise for applications in the coffee industry, it has its limitations. “Specialty coffee must score between 80 and 100, but our model can’t tell whether beans are 80 or 90. That would require machine learning with samples for each score in order to specify these categories in the mathematical model,” study author Winston Pinheiro Claro Gomes said in a news release. He intends to do more work using coffee beans that correspond with each SCA score.
This study isn’t the only one to leverage MSI for the classification of coffee beans. In a recent study published in Food Chemistry: X, researchers used MSI to discriminate between arabica and robusta beans.
MSI has other potential applications in coffee, which researchers have been exploring. “The use of MSI in the coffee industry is very recent. It’s mostly used to map nitrogen in coffee groves, detect necrosis in beans, and detect pests and diseases in plants, as can be seen from the literature on the subject,” Gomes explained in the news release.