Image Generation System for Weld Defect Simulation

Resolve Bottlenecks,
Find Innovative SolutionsGenerate Solutions

Solution Overview

Problem

Existing image generation systems struggle to produce images suitable for machine learning, particularly in generating images that accurately represent defective products for weld inspection.

Innovation Solution

The system includes a first image acquirer, a second image acquirer, and an image processor that performs image transformation and superposition processing. The image transformation processing involves deforming an extracted part of the first image according to the shape of the second object in the second image, and the superposition processing generates a superposed image by combining the transformed image with the second image.

Engineering Contradictions & Design Principles

VSEngineering Contradiction Analysis

1Manufacturing precision

If simple image replacement is used to generate images, then image generation is easy and fast, but the generated images do not accurately represent defective products for machine learning

Engineering Contradiction:

Improveaccuracy of generated defective product imagesVSAvoidcomplexity of image generation process

Core Design Contradiction:

Manufacturing precisionVSDevice complexity

Solution Approach 1:

The patent segments the first image into multiple regions including the defective part, surrounding area, and other components. This segmentation allows selective processing and transformation of specific regions while preserving others, enabling accurate representation of defective products in generated images without unnecessarily complicating the entire image generation process.

Inventive Principle:

Principle #1Segmentation

Solution Approach 2:

The patent introduces an image generation system as an intermediary between actual product images and machine learning training data. This system performs automated transformations including deformation, color adjustment, and composite generation, serving as a mediator that converts real images into realistic defective product images suitable for training without requiring manual image creation.

Inventive Principle:

Principle #24Intermediary (Mediator)

2Manufacturing precision

If image transformation processing is performed to deform extracted parts, then the generated images more closely resemble actual defective products, but the processing time and computational resources increase

Engineering Contradiction:

Improverealism of generated defective product imagesVSAvoidimage generation time

Core Design Contradiction:

Manufacturing precisionVSLoss of time

Solution Approach 1:

The patent performs preliminary actions by pre-processing the first image to extract and segment the defective part and surrounding area before the actual image generation. This preliminary segmentation and extraction of key elements allows for faster transformation and composite generation later, as the critical components are already prepared and isolated for efficient manipulation.

Inventive Principle:

Principle #10Preliminary action

Solution Approach 2:

The patent applies partial transformation processing by selectively deforming and transforming only the extracted defective part and surrounding area, rather than processing the entire image. This partial action approach maintains realism in the critical defective regions while reducing overall computational burden and processing time compared to full-image transformation.

Inventive Principle:

Principle #16Partial or excessive action

3Reliability

If multiple images are superposed to create realistic defective product images, then the quality of training data improves, but the complexity of the generation system increases

Engineering Contradiction:

Improvequality of machine learning training dataVSAvoidcomplexity of image processing system

Core Design Contradiction:

ReliabilityVSDevice complexity

Solution Approach 1:

The patent merges multiple image elements including the transformed defective part, surrounding area, and second image through composite generation. This merging process combines processed components into a unified realistic defective product image, improving training data quality by integrating multiple visual elements while managing system complexity through automated composite generation algorithms.

Inventive Principle:

Principle #5Merging (Combining)

Data Source

PatentUS20250200846A1Image generation system, image generation method, and program

Publication Date: 2025.06.19 PANASONIC INTELLECTUAL PROPERTY MANAGEMENT CO LTD

AI summary

An image generation system includes a first image acquirer, a second image acquirer, and an image processor. The image processor performs image transformation processing and superposition processing. The image transformation processing includes generating, based on a first image, a transformed image by subjecting a predetermined extracted part of a first object to image processing. The superposition processing includes generating a superposed image by superposing the transformed image on a second image. The image transformation processing includes processing of deforming an extracted part according to a shape of a second object shot in the second image.