Image processing-based occluded items detection for self-checkout systems

Resolve Bottlenecks,
Find Innovative Solutions
Generate Solutions

Solution Overview

Problem

Current item checkout systems struggle with inefficient and inaccurate recognition of multiple items due to occlusion, requiring extensive training on numerous images and manual annotation, which is time-consuming and impractical.

Innovation Solution

A method and system utilizing two machine-learning algorithms to identify non-occluded and occluded items by extracting features and defining bounding boxes, reducing the need for exhaustive training data and manual annotation.

Engineering Contradictions & Design Principles

VSEngineering Contradiction Analysis

1Productivity

If multiple items are placed on the checkout counter for recognition, then the checkout process can handle multiple items simultaneously, but items will partially cover or occlude full views of one another, leading to poor recognition accuracy

Engineering Contradiction:
Improvecheckout speedVSAvoiditem recognition accuracy
Core Design Contradiction:
ProductivityVSMeasurement precision

Solution Approach 1:

The patent divides the image processing task into two distinct algorithms: a first algorithm for non-occluded items and a second algorithm for occluded items. This segmentation allows each algorithm to be optimized for its specific task, with the second algorithm specifically trained to handle the challenges of partial occlusion, thereby maintaining high accuracy while processing multiple items simultaneously

Inventive Principle:
Principle #1Segmentation

Solution Approach 2:

The patent applies partial action by using different processing approaches for different items in the image. Non-occluded items are processed by a simpler first algorithm, while only occluded items are processed by the more complex second algorithm. This selective application of processing power maintains accuracy for occluded items without unnecessarily complicating the processing of clearly visible items

Inventive Principle:
Principle #16Partial or excessive action

2Measurement precision

If training is performed on all possible combinations of items covering one another from different positions and angles, then recognition accuracy for occluded items may improve, but the training process becomes infeasible due to the enormous size and manual annotation requirements

Engineering Contradiction:
Improveoccluded item recognition accuracyVSAvoidtraining data complexity
Core Design Contradiction:
Measurement precisionVSDevice complexity

Solution Approach 1:

The patent extracts and isolates the specific challenge of occluded item recognition by creating a separate second machine-learning algorithm dedicated solely to this task. This algorithm is trained specifically on occluded item images, separating the occlusion problem from general item recognition and enabling focused training without requiring exhaustive coverage of all possible item combinations

Inventive Principle:
Principle #2Taking out (Extraction)

Solution Approach 2:

The system performs preliminary classification to identify which items are occluded before applying the complex second algorithm. By first determining which items require special handling, the system prepares the data in advance, allowing the second algorithm to focus only on the relevant occluded items rather than processing all items equally

Inventive Principle:
Principle #10Preliminary action

Data Source

PatentEP3933677B1Image processing-based occluded items detection for self-checkout systems
Publication Date: 2025.11.19 NCR VOYIX CORP
  • EP3933677B1 patent drawingFigure 1
  • EP3933677B1 patent drawingFigure 2
  • EP3933677B1 patent drawingFigure 3

AI summary

Item recognition of a given item is trained on a single item from different views. The item recognition is then trained on images of the given item partially occluded by a second item having same, similar, or different shapes and features to that of the given item. General features of the item are noted and used to detect the given item when the given item is presented with multiple different items having multiple different occluded views.