TOP  >  Thesis/Dissertation  >  Monitoring Grocery Inventory in Refrigerators using Appearance and Load Balance Features

Monitoring Grocery Inventory in Refrigerators using Appearance and Load Balance Features

In this study, we propose a method for monitoring grocery inventory in refrigerators. One of the main contributions is an automated approach for monitoring grocery inventory. When a grocery item enters a refrigerator, the system registers its image, weight and time-stamp to an inventory list. When a grocery item exits from the refrigerator, the system re-identifies it by matching it to an entry in the inventory list and deletes the entry. In this approach, re-identification of groceries has a technical difficulty because there exist various kinds of groceries and they are stored in various ways in the refrigerator. The use of two features for this grocery re-identification, which is the other main contribution of this study, enables automated re-identification. One is Appearance Feature, which reflects the appearance of a grocery item, and the other is Load Balance Feature, which reflects the weight of the grocery item and the position of it on a shelf in the refrigerator. In the experiments, twenty grocery items were stored in and taken out of the shelf in the refrigerator. The result shows all the grocery items are correctly re-identified by the proposed method.