Real Time Pharmaceutical Product Recognition Using Color and Shape Indexing

In this paper, a real time application for recognition of pharmaceutical product boxes is presented. The system developed is able to recognize in less than a second a box from a set of 10000 pharmaceutical products using as cues the box size, an optimal description of a pose invariant representation of the shape and its color. To deal with the huge amount of product database as well as to achieve the real-time performance, the identification method has been designed in a hierarchical way. The inspected product box is classified in one of 50 size clusters, then, the shape and color descriptors are applied to identify the product inside a cluster in a way that the learning and recognition step can be done in an efficient manner. Additionally, principal component analysis is applied to reduce the dimensionality of the identification problem of the product boxes. The system is being currently validated and in use in the biggest pharmaceutical store serving most of the pharmacies in Cataluña, Spain.