Quality in beans: tracking and tracing coffee through automation and machine learning

Abstract

Purpose

Coffee is one of the most consumed beverages in the world and the global coffee industry is worth over $100bn. However, the industry faces significant sustainability challenges. Developing a quality traceability system to select the coffee beans and to ensure their authentication would result in economic advantages, because it allows for fraud to be avoided and increases consumer confidence.

Design/methodology/approach

Traceability is one of the key elements of sustainability in the coffee sector. The literature reveals that near-infrared (NIR) approaches have a huge potential for gaining rapid information about the origin and properties of coffee beans, without invasive procedures. This study demonstrates the scalability potential of automated methods of manipulation and image acquisition of coffee beans, from experimental scale to industrial lines.

Findings

A solution based on the interaction of a manipulation system, a NIR spectrometer acquisition station integrated with a machine learning infrastructure and a compressed air classifier allows for the automatic separation of coffee beans into different classes of origin.

Originality/value

Apart from traceability, the wide industrialization of this system offers further advantages, including reduced workforce, decreased subjectivity in the evaluation and the acquisition of real-time data for labeling.