Store products recognition and counting system using computer vision

Abstract

The aim of this study is to recognize products in a store shelves image using Speed Up Robust Features (SURF) and color histogram. This combination helps to provide more accuracy in categorizing the products to help the owners to avoid problems like out of stock and products misplacement. The results of the detection are stored in a database to make in much easier and faster to process this information later in order to create a custom service as requested by the owners. The accuracy of the used algorithm is demonstrated using two scenarios, the first scenario uses one model image for each product while the second one uses three model images for each product. The results illustrate a huge improvement in the results accuracy by providing more model images for each product.