Digital Printing Defect Monitoring With Synthetic Training Images
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
The generation of a training database for neural networks to detect print defects in digital printing processes is inefficient and time-consuming due to the rarity of defects, requiring extensive manual image acquisition over long periods.
Innovation Solution
A method to train neural networks using digitally manipulated images to introduce print defects, allowing rapid construction of a comprehensive training database without manual operations, by superimposing, blurring, or altering digital images to replicate various defects.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If manual image acquisition is used to build training database, then defect detection accuracy is improved, but time consumption and resource usage increase significantly
Solution Approach 1:
The patent creates synthetic training images by copying and manipulating digital images of substrates and print patterns, then adding simulated defects through digital processing. This copying approach eliminates the need for manual acquisition of real defective prints while maintaining training effectiveness.
Solution Approach 2:
The patent performs preliminary digital manipulation of images to introduce defects before training the neural network. By pre-processing images to include various defect types (stains, blurs, geometric distortions), the system prepares comprehensive training data in advance without waiting for natural defect occurrence.
2Adaptability or versatility
If manual image acquisition is used to build training database, then comprehensive defect coverage is improved, but productivity decreases
Solution Approach 1:
The system generates diverse training examples by copying base images and applying various digital transformations to simulate different defect types and conditions, achieving comprehensive defect coverage without manual collection.
Solution Approach 2:
The patent systematically varies parameters such as defect position, size, shape, and type through digital manipulation of training images. This parameter variation approach ensures comprehensive defect coverage across multiple dimensions while maintaining high productivity through automated processing.
3Reliability
If manual operations are used for training data preparation, then data quality is improved, but energy consumption increases
Solution Approach 1:
The patent replaces manual mechanical operations (physical image acquisition, handling, and processing) with digital computational operations. This substitution maintains data quality through systematic digital generation while dramatically reducing energy consumption associated with physical operations.
4Adaptability or versatility
If extensive manual acquisition is performed to capture rare defects, then defect variety is improved, but loss of substance increases
Solution Approach 1:
The system creates virtual representations of defective prints through digital copying and manipulation, eliminating the need to produce and discard large quantities of physical substrate and ink to capture rare defect instances.
Solution Approach 2:
The patent converts the limitation of rare natural defect occurrence into an advantage by using digital manipulation to intentionally create and control defect instances. This transforms the harmful waste associated with extensive production into a beneficial automated process that generates diverse training data without physical resource consumption.
Data Source
Figure 1
Figure 2
Figure 3
AI summary
A method for training a neural network model (320) for use in print defect monitoring applications in a substrate printing process. The method comprises, under the control of a computing system (180): - providing (630) to the computing system a plurality of digital images to be printed; - providing (626) to the computing system one or more substrate digital images each representative of a corresponding substrate; - generating (650), by the computing system, a plurality of first training images representative of defect-free prints, said generating the first training images comprising combining (668) each of a group of said digital images to be printed with one of the substrate digital images; - generating (650), by the computing system, a plurality of second training images representing prints each containing at least one print defect, said generating the second training images comprising carrying out the following operations a), b) on each of a group of said digital images to be printed: a) digitally manipulating (660) said digital image to be printed to introduce into it at least one graphic alteration representing the corresponding at least one print defect; b) combining (668) the digitally manipulated digital image to be printed with one of the substrate digital images; - training (604), by the computing system, said neural network model on the basis of said first training images and said second training images to optimize the capability of said neural network model to classify the first training images as defect-free, and to classify the second training images as containing at least one defect.