Coffee Bean Inspection
Classify coffee beans to ensure the right varieties are used in blends
Coffee roasters and distributors often blend different coffee bean varietals to create a specific and consistent flavor profile. These blends are formulated out of as many as five different specific varietals of beans, and vendors compete by creating new and distinct blends.
Batches of beans arrive from different vendors and need to have their identity confirmed before being inventoried. Including the wrong bean type can change the flavor profile of the final product and lead to consumer dissatisfaction. Each incoming batch or batch on its way for blending must also be checked for physical contaminants such as plant matter, pebbles, and other material.
Coffee beans within a single varietal can differ significantly in appearance, while the overall differences between varietals can depend on many different parameters, including size, shape, surface texture, and color. Conventional machine vision has trouble accepting the variation within a bean type while making clear decisions between types and therefore cannot accurately solve visual coffee bean inspection applications.
Cognex Deep Learning is ideal for coffee bean sorting because it can easily and quickly discern between different colors, sizes, and textures of beans. It trains on sets of images of each bean type that will be arriving at inventory. The classification tool distinguishes between types while accepting natural variation within each type, even when different types are visually similar. The defect detection tool notes any physical impurities before a batch of beans goes into the blend.
If a new bean type is introduced to create new blends, all that is required is a short training on an image set of that varietal for it to be properly identified and sorted, without any programming necessary.