Global and local category-based item location filtering

Abstract

A system and method for global and local category-based item location recognition filtering is provided. A plurality of candidate items associated with a current location within a retail environment are filtered using a global level filter that identifies and removes items associated with an item category that is not a confirmed category for the recognized location. A local level filter analyzes item count values for the remaining candidate items to identify and remove low-count items. If any single item count is below a threshold, the low count items are removed. If all the candidate items are low count items at the item level, a count values for all candidate items in the same category are aggregated into a category-level item count value which is used to identify low count items. The low-count items are filtered out to eliminate false positives for improved accuracy and reliability of the item-to-location mapping results.