Automated classification of starch granules using supervised pattern recognition of morphological properties

Image analysis techniques have been used to investigate the likelihood of being able to classify and assign a probability regarding the plant origin of individual starch granules in a collection of granules. Quantifiable variables were used to characterize the granules, and the assignments and probabilities were calculated objectively. We consider the classification of images containing granules of a single species and of mixed species and the possibility of assigning a class to granules of unknown species in an image of a slide obtained from the dental calculus of chimpanzees.