Production Print Inspection Using OCR to Cut False Positives
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current optical character recognition (OCR) systems for continuous inkjet (CIJ) in industrial applications have low reliability due to high rates of false positives, leading to costly manual inspections and production disruptions, and are expensive compared to cameras without vision systems.
Innovation Solution
A multi-stage production print inspection system that includes a printer device interfaced with a camera and a computing device, utilizing trained OCR algorithms and image recognition algorithms to validate printed content on various substrates, with a second inspection stage involving human intervention for accuracy, and machine learning to improve performance.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If a vision system with smart camera is used for print inspection, then measurement precision is improved, but device complexity and cost increase
Solution Approach 1:
The inspection system is divided into two distinct stages: a first automated stage using machine learning for initial inspection, and a second manual stage for verification of borderline cases. This segmentation allows the system to achieve high accuracy without requiring a single complex vision system to handle all cases, thereby reducing overall device complexity while maintaining measurement precision.
Solution Approach 2:
A multi-classification model serves as an intermediary between simple camera capture and complex vision system analysis. The model pre-processes images by identifying print quality issues and assigning priority levels, which then guides the second-stage manual inspection. This intermediary layer reduces the complexity burden on the vision system while maintaining high inspection accuracy.
2Measurement precision
If a vision system with smart camera is used for print inspection, then measurement precision is improved, but cost increases
Solution Approach 1:
The inspection workflow is segmented into automated low-cost image capture and analysis, followed by selective manual inspection. By dividing the inspection task, the system achieves high measurement precision without requiring expensive smart cameras with built-in vision processing for every unit, thereby reducing overall system cost while maintaining accuracy.
Solution Approach 2:
The system uses standard, inexpensive cameras rather than expensive smart cameras with integrated vision systems. The inspection intelligence is moved to software (multi-classification models) rather than hardware, allowing the use of cheaper camera components while achieving the same measurement precision through algorithmic processing.
3Productivity
If automated vision inspection is used, then productivity is improved, but reliability deteriorates due to false positives
Solution Approach 1:
The inspection process is segmented into two stages: automated ML-based screening for high-speed initial assessment, followed by manual verification for borderline cases. This segmentation maintains high productivity through automation while improving reliability by having human inspectors confirm ambiguous cases, thereby reducing false positives.
Solution Approach 2:
The system implements feedback loops where manual inspection results from the second stage are used to refine and retrain the multi-classification models. This continuous feedback improves the reliability of automated inspection over time by reducing false positives, while maintaining high productivity through the automated first stage.
4Productivity
If automated vision inspection is used, then productivity is improved, but loss of time increases due to manual re-inspection
Solution Approach 1:
The inspection is segmented so that clear-cut cases are resolved quickly by automated ML analysis, while only ambiguous borderline cases proceed to manual inspection. This segmentation minimizes the overall time loss by limiting manual re-inspection to only the necessary subset of cases, thereby maintaining high productivity while reducing unnecessary manual time consumption.
Data Source
Figure 1a
Figure 1B
Figure 2~3
AI summary
A device comprising a printer configured to apply a code of printed content on a substrate of a product based on a printer technology type, the code having a plurality of digits. The device includes an optical code detector, executed by one or more processors, to detect the code in a received image of the product printed by the printer by optically recognizing characters in the received image using a trained optical character recognition (OCR) algorithm for the printer technology type. The OCR algorithm is trained to identify each digit of the plurality of digits of the code in a region of interest (ROI) based on at least one product parameter to which the printed content is directly applied and the printer technology type. A system and method are also provided.