The recognition system for zip-top cans in the beverage showcase

The beverage showcase that is used to store cold drinks is ubiquitous in supermarkets, which have brought great convenience. However, a great disadvantage of the beverage showcase is that it needs manual check of remaining volume. Therefore, a recognition system is proposed in this paper to solve this tricky problem and to make our lives more intelligent. Positions, brands and the number of zip-top cans could be identified easily in the recognition system which monitors the dynamics of beverages in the beverage showcase. After starting the system, the key points detector that matches input picture and feature dataset is first executed. It will recognize zip-top cans that are less partially occluded or not covered. Next, the system carries out the rims detector to detect all upper rims of zip-top cans, which determine positions and the number of cans. In the rims detector, we propose an evaluation algorithm for the qualified ellipse. Then the colors detector detects main color blocks near the upper rims of zip-top cans to determine brands of cans covered each other. Three detectors are combined to constitute a recognition system with high accuracy and practical application value. In the system, the recognition rate is above 85%.