Retail Item Identification Using Shelf-Based Candidate Lists
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing image processing systems for retail environments face inefficiencies in identifying items from large dictionaries, leading to prolonged matching times, especially in stores with a wide variety of items.
Innovation Solution
An image processing apparatus and server device configuration that includes a network interface, processor, camera, and user interface, which captures images of items in shopping baskets, compares them to a list of items received from the server, and uses pattern matching techniques to quickly identify items by generating a candidate list based on user access to specific shelf divisions.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Adaptability or versatility
If a large dictionary of item images is used to cover all possible items in a retail store, then the system can identify any item, but the matching process takes much time
Solution Approach 1:
The patent segments the item identification process into two stages: first, use shelf division information to narrow down to a small candidate list of likely items; second, perform pattern matching only within this reduced candidate list. This segmentation dramatically reduces the effective dictionary size for matching while maintaining comprehensive item identification capability across the entire store.
Solution Approach 2:
The system performs preliminary action by pre-storing the correspondence between shelf divisions and item lists in a database. Before the actual matching process, the system retrieves the relevant item list based on the captured shelf division information, preparing a pre-filtered candidate list that reduces the number of items requiring pattern matching.
2Measurement precision
If the system stores comprehensive dictionary information for all items, then item identification is thorough, but the device complexity increases
Solution Approach 1:
Instead of storing a single comprehensive dictionary of all items, the system segments the dictionary into multiple item lists organized by shelf division. Each shelf division has its own associated item list, allowing the system to retrieve only the relevant subset for matching based on the captured shelf division, thereby reducing the effective dictionary size while maintaining identification accuracy.
Solution Approach 2:
The patent introduces shelf division information as an intermediary that bridges the gap between the captured image and the item dictionary. The shelf division acts as a mediator to select the appropriate item list from the database, enabling the system to access comprehensive item information only when needed through this intermediate filtering mechanism.
Data Source
AI summary
An image processing apparatus includes a network interface configured to communicate with a server device, a first camera, and a processor. The processor is configured to identify an item presented by a user and imaged by the first camera using a list of items received from the server device through the network interface, wherein the items in the list are displayed at a location accessed by the user.