Printing Medium Identification via Spectral Learning Model

Resolve Bottlenecks,
Find Innovative Solutions
Generate Solutions

Solution Overview

Problem

Existing methods for determining the type of printing medium struggle to accurately classify media due to insufficient differences in color value characteristics, making it difficult to add new categories and properly process unknown media.

Innovation Solution

A printing medium processing system that uses machine learning to identify media by analyzing spectral information from primary colors and paper white areas, employing a learned model to determine the type of printing medium and adjust printing conditions accordingly, with the option to update the model through transfer learning for unknown media.

Engineering Contradictions & Design Principles

VSEngineering Contradiction Analysis

1Measurement precision

If color value parameters are classified in advance to determine medium type, then medium identification can be performed, but it becomes difficult to properly classify media and add new categories when color value differences are insufficient

Engineering Contradiction:
Improvemedium identification accuracyVSAvoidability to add new medium categories
Core Design Contradiction:
Measurement precisionVSAdaptability or versatility

Solution Approach 1:

The patent changes the parameter used for medium identification from simple color values to spectral information, which provides more detailed and distinguishable characteristics across different wavelengths. This enables both accurate identification of known media and differentiation of new media types with subtle differences.

Inventive Principle:
Principle #35Parameter changes

Solution Approach 2:

The patent transitions from one-dimensional color value parameters to multi-dimensional spectral information across multiple wavelengths, adding dimensional depth to the identification process. This allows for more nuanced differentiation between media types and facilitates the addition of new categories.

Inventive Principle:
Principle #17Another dimension (Dimensionality change)

2Productivity

If preset parameters are used for medium classification, then processing can be performed according to identified medium type, but media with insufficient characteristic differences cannot be properly distinguished

Engineering Contradiction:
Improveprocessing efficiencyVSAvoidmedium differentiation capability
Core Design Contradiction:
ProductivityVSMeasurement precision

Solution Approach 1:

The system replaces preset color value parameters with spectral information parameters that capture more distinctive characteristics of different media. This enables both efficient automated processing and precise differentiation of media with subtle differences in their optical properties.

Inventive Principle:
Principle #35Parameter changes

3Measurement precision

If machine learning with spectral information is used for medium identification, then accurate identification without special parameters is achieved, but computational complexity increases

Engineering Contradiction:
Improvemedium identification accuracyVSAvoidcomputational complexity
Core Design Contradiction:
Measurement precisionVSDevice complexity

Solution Approach 1:

The patent performs machine learning training in advance to create a pre-trained model that captures medium identification patterns. During actual operation, the system applies this pre-trained model rather than performing full training, significantly reducing computational complexity while maintaining high identification accuracy.

Inventive Principle:
Principle #10Preliminary action

Solution Approach 2:

The system extracts only the essential spectral information features needed for identification rather than processing all possible spectral data. This selective extraction reduces computational complexity while preserving the accuracy needed for reliable medium identification.

Inventive Principle:
Principle #2Taking out (Extraction)

Data Source

PatentUS11763118B2Printing medium processing system and printing medium processing method
Publication Date: 2023.09.19 SEIKO EPSON CORP
  • US11763118B2 patent drawing
  • US11763118B2 patent drawing
  • US11763118B2 patent drawing

AI summary

For a plurality of kinds of printing media, spectral information is produced from results of printing of at least five points in paper white of the printing media, black, and three primary colors of coloring and printing materials, and a learned model is generated in advance by machine learning that uses the spectral information as learning data. An identification section that identifies the kinds of the printing media by using the learned model is provided. The kinds of the printing media are identified by applying the spectral information produced from the results of printing of at least five points on a printing target medium on which printing is to be performed to the learned model, and the printing target medium is processed in accordance with the identified kind of the printing target medium. In the case of a new unlearned printing medium, transfer learning is performed to learn the printing medium, and the model is updated.