Authors:
Kirk Goldman, Manda Miller, Kristen Chung, Sam Champagne, Pranit Kotgire
Abstract:
This disclosure proposes a method that allows customers to correct some unintentional errors at Self Checkout without requiring a Retail Store attendant. This new Computer Vision functionality would observe unintentional errors by the shopper that do not result in loss scenarios and guide the shopper into correcting those errors.
Background:
Shoppers often make simple mistakes at self-checkout that require a human intervention but could easily be resolved through use of an AI/CV solution. This solution is automatically enabled when a self-checkout solution sees a scan in the transaction log that does not match what the AI observes. Like an overhead loss prevention solution, the UX asks the shopper if they intended to do what happened (e.g., scan something twice when one item was observed getting put into the bagging area) and then, when answered/corrected, resolves the mistake in the transaction without requiring manual intervention from an attendant. This helps keep inventory accurate, optimizes self-checkout attendant productivity, improves Self-Checkout utilization (speed of transaction) and keeps the shopper happy.
Description:
To do this successfully the solution needs to control both the AI/CV and the UX and operate in real-time (at the EDGE).
Front-end attendant is often called over for minor issues that AI/CV could help a shopper self-adjust.
AI/CV focus has been on loss prevention, but can also assist common shopper mistakes.
Shoppers are generally slower in completing transactions (versus cashiers/attendants) due to periodic interventions. This will speed up turnover at the SCO lanes.
Examples of shopper mistakes include multiple scans, item look ups, etc.
This proposal would be a new and novel function of the ELERA Security Suite. The new Computer Vision functionality would observe unintentional errors by the shopper that do not result in loss scenarios. Specifically, it would continue to monitor shopper and item movement along with the scanning activity at the lane but now will match transaction log (T-log) to see if the shopper has incorrectly scanned items.
In Scenario A, a shopper scans an item twice but only puts down the single item in the bagging area (two passes at the scanner are observed by the CV, two of the same item appear in the T-log, one item was put in the bagging areas.)
- Like a potential LP situation, a UX-based intervention pops up to help the shopper correct the mistake.
- The intervention screen asks the shopper if they intended to scan the item twice while showing the relevant snippet of video.
- The shopper is prompted to answer a question and then repeat the scanning to correct the mistake. Once corrected, the UX goes back into normal shopper mode where they can continue with their self-checkout transaction. This is completed without intervention from an attendant.
In Scenario B, a shopper scans the wrong barcode on an item.
- A shopper scans an item with barcodes for both "eaches" (individual items) and cases. Example: 12 pack of Coke, there is a barcode for the pack and a barcode on each Coke. The AI/CV monitors the shopper and the T-log.
- The solution acts when the CV observes the scan but notices an individual item has been added to the T-log when it should have added a case. Rather than creating a manual intervention or having the shopper ask for help the solution identifies the mistake.
- The solution recognizes the mistaken barcode, it automatically corrects the t-log to the case SKU versus the each SKU.
- If the solution does not recognize the item, an AI generated intervention in the UX helps the shopper re-scan the correct barcode, again without the help of an attendant.
TGCS Reference 00599
