With evolving consumption trends, the coffee market is experiencing increasing demand for high-quality, traceable coffees, which, in turn, has led to price growth. Therefore, due to its increased economic value, coffee has become a constant target of fraudulent actions. As result, many analytical techniques have been explored as tools for coffee classification and authentication, of which the use of digital, hyperspectral and/or multispectral imaging is noteworthy. This type of analysis provides rapid, non-destructive, environmentally friendly, and increasingly accessible alternatives to conventional analytical methods. By consulting three different databases, this work systematically revised articles published in the last 10 years, which utilize digital image analysis and hyper/multispectral imaging for the botanical and geographical classification and authentication of coffees. The reviewed studies (n = 17) demonstrate that, when paired with classification algorithms, discrimination across species, origins, and quality categories can be achieved. A critical point to highlight is the importance of using whole beans and standardizes roast degree to avoid biasing the models. Concerning digital images, relying solely on color features limits the robustness of the classification models. Incorporating complementary textural and shape features is thus necessary to capture the coffee botanical or geographic information, as shown in a minor number of the selected studies. In a similar fashion, for hyper/multispectral imaging, there is still potential to further exploit the spatial information, thus achieving the technique’s full potential. The evidence indicates that image-based methods are steadily progressing into reliable tools for coffee authentication.
Tessaro et al. (Sun,) studied this question.