Multi-Signal Bulk Item Recognition for Occlusion-Resistant Checkout
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Conventional self-checkout and cashier-assisted checkout processes are cumbersome for consumers with large numbers of items due to linear item identification, which requires serial scanning or input, and computer vision-based systems face challenges with occlusion and accuracy.
Innovation Solution
A hybrid checkout system combining computer vision with multiple sensors, including overhead cameras and RFID, captures images and data from multiple angles and positions to identify items in motion, using machine learning for accurate bulk item recognition.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Ease of operation
If linear item identification process is used, then each item can be identified individually, but the process becomes cumbersome for consumers with large numbers of items
Solution Approach 1:
The patent combines multiple detection methods (optical detection, RFID detection, weight detection) into a unified bulk item identification system. Instead of processing items one by one through linear scanning, the system simultaneously detects multiple items using sensor arrays, merging detection operations to achieve both high speed and ease of operation for consumers with large baskets
Solution Approach 2:
The patent replaces manual mechanical scanning operations with automated sensor-based detection systems. Optical sensors, RFID readers, and weight sensors automatically identify items without requiring consumers to manually scan or enter each item, substituting mechanical manual operations with automated detection to improve both productivity and ease of operation
2Productivity
If computer vision-based systems are used for bulk item recognition, then transaction velocity can be improved, but accuracy is compromised due to occlusion
Solution Approach 1:
The patent creates a multi-functional detection system that combines optical detection, RFID detection, and weight detection. Each sensor type serves a specific function: optical sensors capture visual data for item identification, RFID readers provide precise item tracking, and weight sensors detect item presence and mass. This multi-functional approach compensates for the limitations of any single detection method, maintaining both high transaction velocity and accurate item recognition even in challenging conditions like occlusion
3Measurement precision
If multiple sensors are deployed for bulk item identification, then recognition accuracy is improved, but device complexity increases
Solution Approach 1:
The patent segments the detection system into distinct functional modules: optical detection units, RFID detection units, and weight detection units. Each module operates independently and processes specific types of data, allowing the system to handle complexity through modular organization. The segmentation enables independent optimization of each sensor type while maintaining overall system coherence, improving accuracy without overwhelming complexity
4Device complexity
If linear item identification is used, then simple processing is maintained, but operator burden increases for consumers with large baskets
Solution Approach 1:
The patent implements self-service detection where the automated sensor system performs item identification without requiring consumer intervention. The system automatically detects items as they are placed in the basket, eliminating the need for consumers to manually scan or enter each item. This self-service approach dramatically reduces operator burden for consumers with large baskets while maintaining relatively simple processing through automated multi-sensor detection
Applied Scientific Principles
This section explains which scientific principles are used to turn an abstract innovation direction into a practical engineering solution.
Function Achieved in This Case
The hybrid system enhances transaction velocity and accuracy by transforming the linear item identification process into a bulk process, reducing operator burden and improving recognition confidence through multi-signal data analysis.
Implementation Method 1
overhead cameras located at different positions on the computer vision apparatus and having at least partially non-overlapping fields-of-view (FOVs)... capture images of the items from multiple different angles/vantage points
Implementation Method 2
one or more RFID sensors... additional sensor data (e.g., RFID data) captured by one or more RFID sensors
Data Source
AI summary
A hybrid checkout system that enables in-motion, multi-signal, bulk item identification is disclosed. The hybrid checkout system includes a computer vision apparatus that includes a plurality of cameras fixed at different locations relative to a conveyor belt. The conveyor belt includes markings that assist a user with item placement. RFID sensors are provided with the hybrid checkout apparatus. The cameras capture multiple images of items placed on the belt as the items are in motion. Each camera captures multiple images of the items captured from different vantage points as the items are in motion. In addition, the RFID sensors gather RFID data from RFID tags affixed to the items, as the items are in motion. The image data and other sensor data is provided as a multi-signal input to a machine learning model that is trained to recognize items and output item identifiers for the items. Pricing information corresponding to the item identifiers received from the model is determined and the items and their respective prices are added to a transaction record.