Self-Checkout Cart Validation Using CNN Bagging Area Item Counts
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Customers scanning multiple items at self-checkout machines often introduce errors by scanning items multiple times or missing scans, leading to discrepancies in inventory and charges, necessitating manual intervention.
Innovation Solution
Integrate a camera with a self-checkout machine to monitor the bagging area, using a convolutional neural network (CNN) to count items accurately and adjust the shopping list or alert the customer to discrepancies.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Productivity
If customers scan one item multiple times before placing all items in the bagging area, then the checkout process becomes faster and more convenient, but errors such as overcharge or undercharge occur due to mismatch between scanned quantity and actual items
Solution Approach 1:
The system continuously captures images of the bagging area and compares the detected item quantity with the scanned item quantity in real-time. When a discrepancy is detected, the system provides feedback by displaying an alert message to the customer, prompting them to verify the bagging area. This closed-loop feedback mechanism maintains transaction accuracy while allowing customers to scan items efficiently without immediate verification.
Solution Approach 2:
The patent replaces the traditional mechanical verification process (manual counting and verification by cashiers) with an automated computer vision system using convolutional neural networks. The CNN model automatically detects and counts items in the bagging area, substituting human labor with an automated optical detection system that operates continuously without interrupting the checkout flow.
2Reliability
If manual verification is implemented to check scanned items against bagged items, then transaction accuracy improves, but checkout time increases and customer convenience decreases
Solution Approach 1:
The system performs preliminary detection by continuously capturing images of the bagging area and pre-calculating the item quantity before the customer completes scanning. The comparison between scanned quantity and detected quantity is performed in advance, allowing the system to prepare alert messages ready for display. This preliminary action eliminates the need for post-scanning verification delays.
Solution Approach 2:
The system enables customers to self-verify by displaying alert messages that prompt them to check their bagging area. Instead of requiring cashier intervention, the automated system empowers customers to independently verify their own transactions by comparing the alert message with the actual items in the bagging area, maintaining accuracy while preserving customer autonomy and checkout speed.
3Device complexity
If traditional scanning-only systems are used, then device complexity remains low, but inventory management discrepancies occur due to undetected scanning errors
Solution Approach 1:
The patent introduces an intermediary computer vision system that acts as a mediator between the scanning system and inventory management system. The CNN-based detection system captures images of the bagging area and provides an independent verification layer, comparing detected items with scanned items. This intermediary layer prevents inventory discrepancies without requiring fundamental changes to the existing scanning infrastructure.
Solution Approach 2:
The system creates a visual copy of the bagging area through image capture and processing. By generating a digital representation (image) of the physical items and analyzing it through CNN, the system obtains a copy-based verification of item quantity that complements the scanning data. This copying approach provides inventory verification without requiring direct physical manipulation or complex sensor integration.
Data Source
AI summary
The present disclosure provides techniques for tracking and verifying items that are scanned multiple times in a self-checkout system. A plurality of inputs indicating more than one scan of an item are received at a self-checkout machine. An image of a bagging area is captured. A CNN is trained to recognize the item based on a training dataset and a selected training algorithm. One or more units of the item are detected in the bagging area from the image using the CNN. A number of the detected units of the item in the bagging area is compared with a number of times the item has been scanned. Upon determining that there is no match, a shopping list displayed on the self-checkout machine is adjusted to reflect the number of the detected units of the item in the bagging area.