This article describes the methodology used for the automatic classification of finished products at Familia Sancela Company, Medellin plant, by visual recognition of numeric codes labels, printed on their packaging, before the stowage and storage procedures. Based on the morphology and package design and techniques using digital image processing and artificial vision, it seeks to graphically detect a numeric label that encodes the product, whose characters are framed in a box. For this, an image preprocessing by thresholding, are the outlines of the image and using the polynomial approximation method detected the rectangle that frames the numerical code, this region is applied an orientation correction algorithm, it is a segmentation of each digit in individual images and finally apply the algorithm of Optical Character Recognition (OCR), which determines the value of the character by comparing the Euclidean distances between the projection of the character and the established databases. The implementation of this automation results in an optimization in the packaging procedure as well as decrease of time, costs and errors. All processing is done using the computer vision library, OpenCV and cvBlobsLib, in the development platform Microsoft Visual Studio C + + 2010.