Machine Vision Parameter Estimation Using Multiple Examples

Resolve Bottlenecks,
Find Innovative Solutions
Generate Solutions

Solution Overview

Problem

Current machine vision systems require manual and time-consuming refinement of object recognition models, leading to suboptimal performance and increased execution times due to the use of single-image or CAD-based parameter estimation, which can result in unstable model points and inefficient search processes.

Innovation Solution

A method for automating the refinement of object recognition models by using multiple images to identify stable model points and estimating level-specific parameters, including combining and decision functions to update model contours and parameters, thereby improving model robustness and accuracy.

Engineering Contradictions & Design Principles

VSEngineering Contradiction Analysis

1Measurement precision

If manual refinement of object recognition models is performed, then model accuracy can be improved, but time consumption and complexity increase significantly

Engineering Contradiction:
Improvemodel accuracyVSAvoidrefinement time
Core Design Contradiction:
Measurement precisionVSLoss of time

Solution Approach 1:

The system automatically refines model parameters by processing multiple example images and computing optimal values without user intervention. The refinement process serves itself by using the input images to generate improved model parameters, eliminating the need for manual user refinement while maintaining or improving accuracy.

Inventive Principle:
Principle #25Self-service

Solution Approach 2:

The system performs preliminary refinement by processing example images before actual object recognition tasks. By pre-computing optimal model parameters from example images, the system prepares the model in advance, avoiding time-consuming manual refinement during operational use.

Inventive Principle:
Principle #10Preliminary action

2Ease of manufacture

If single-image parameter estimation is used, then setup process is simple, but model robustness and detection accuracy deteriorate

Engineering Contradiction:
Improvesetup simplicityVSAvoidmodel robustness
Core Design Contradiction:
Ease of manufactureVSReliability

Solution Approach 1:

The system uses multiple example images to estimate model parameters, making the parameter estimation process more universal and adaptable to various lighting conditions, object variations, and environmental factors. This multi-image approach creates a more robust model that performs reliably across different scenarios while maintaining ease of setup through automated processing.

Inventive Principle:
Principle #6Universality (Multi-functionality)

Solution Approach 2:

The system combines information from multiple example images to generate optimized model parameters. By merging data from multiple sources, the system creates a more comprehensive and robust model representation that captures various object appearances and conditions, improving reliability without complicating the setup process.

Inventive Principle:
Principle #5Merging (Combining)

3Measurement precision

If manual model parameter optimization is performed, then detection accuracy can be improved, but execution time increases due to complex processing

Engineering Contradiction:
Improvedetection accuracyVSAvoidprocessing speed
Core Design Contradiction:
Measurement precisionVSProductivity

Solution Approach 1:

The system performs parameter optimization in advance by processing example images before actual detection tasks. The computationally intensive optimization is done preliminarily during model creation, allowing fast execution during operational use without sacrificing detection accuracy.

Inventive Principle:
Principle #10Preliminary action

Solution Approach 2:

The system creates an optimized model representation by processing example images and generating refined parameters that capture essential object characteristics. This optimized model copy can then be used for fast detection without requiring repeated complex processing of original images.

Inventive Principle:
Principle #26Copying

4Extent of automation

If automated parameter estimation is implemented, then user interaction is minimized, but parameter optimality may be compromised without expert tuning

Engineering Contradiction:
Improveautomation levelVSAvoidparameter optimality
Core Design Contradiction:
Extent of automationVSMeasurement precision

Solution Approach 1:

The automated parameter estimation system serves itself by using multiple example images to compute optimal parameters without requiring expert user intervention. The system independently analyzes the example images and generates optimized model parameters, achieving both high automation and parameter optimality simultaneously.

Inventive Principle:
Principle #25Self-service

Solution Approach 2:

The system uses feedback from multiple example images to iteratively refine parameter estimates. By processing multiple examples and using their collective information, the automated system achieves parameter optimality that would otherwise require expert tuning, while maintaining full automation.

Inventive Principle:
Principle #23Feedback

Data Source

PatentEP4636707A1Methods and systems and automatic example-based parameter estimation in machine vision
Publication Date: 2025.10.22 MVTEC SOFTWARE
  • EP4636707A1 patent drawingFigure 1~2
  • EP4636707A1 patent drawingFigure 3
  • EP4636707A1 patent drawingFigure 4

AI summary

The present disclosure provides methods that estimate or improve various parameters of an object recognition model to improve runtime, accuracy and robustness while minimizing the required user interaction to optimize these values. In one aspect of the invention, methods are defined to generate object recognition models with refined contours. A further method is defined by the invention to estimate level-specific parameters for the object recognition algorithms.