Classification Model Training via Operator Feedback

Resolve Bottlenecks,
Find Innovative Solutions
Generate Solutions

Solution Overview

Problem

Current methods for assigning classes of interest within measurement data in surveying and metrology are time-consuming and require significant human intervention, especially when dealing with unstructured and inhomogeneous datasets, and existing machine learning solutions face challenges in adapting to diverse applications and environments.

Innovation Solution

A method that acquires and prepares suitable training data for each measurement task, utilizing feedback from operators to improve classifier training, reducing data storage and overhead, and implementing a notification functionality for uncertain classifications to reduce human intervention and improve classification accuracy.

Engineering Contradictions & Design Principles

VSEngineering Contradiction Analysis

1Productivity

If machine learning algorithms are used for automated classification, then productivity is improved, but device complexity increases due to computational challenges and data processing requirements

Engineering Contradiction:
Improveclassification speedVSAvoidcomputational complexity
Core Design Contradiction:
ProductivityVSDevice complexity

Solution Approach 1:

The system performs preliminary actions by acquiring and preparing training data in advance for each measurement task. This pre-processing step includes collecting measurement data, generating corresponding classification labels, and storing them as training datasets before the actual classification operation, thereby reducing computational complexity during runtime while maintaining high productivity

Inventive Principle:
Principle #10Preliminary action

Solution Approach 2:

The system implements feedback mechanisms where classification results are evaluated and used to improve the classification model iteratively. Measurement data and corresponding labels are continuously fed back to retrain and refine the machine learning algorithms, enabling the system to adapt to different applications and environments while maintaining efficient automated classification

Inventive Principle:
Principle #23Feedback

2Measurement precision

If manual classification processes are used, then measurement precision is maintained through human judgement, but loss of time increases due to time-consuming processes

Engineering Contradiction:
Improveclassification accuracyVSAvoidprocessing time
Core Design Contradiction:
Measurement precisionVSLoss of time

Solution Approach 1:

The system enables self-service by automatically performing classification tasks using machine learning algorithms without requiring continuous human intervention. The automated classification system processes measurement data independently, generating classification results based on trained models, thereby eliminating time-consuming manual processes while maintaining consistent classification accuracy

Inventive Principle:
Principle #25Self-service

Solution Approach 2:

The system replaces manual mechanical classification processes with automated computational methods. Machine learning algorithms substitute human judgement and manual sorting operations, processing classification tasks computationally to achieve both high accuracy and speed, eliminating the time loss associated with manual classification

Inventive Principle:
Principle #28Mechanics substitution (Replace mechanical system)

3Ease of operation

If generic classification models are used, then ease of operation is improved, but adaptability decreases for diverse applications and environments

Engineering Contradiction:
Improveease of useVSAvoidapplication-specific adaptation
Core Design Contradiction:
Ease of operationVSAdaptability or versatility

Solution Approach 1:

The system implements dynamic adaptability by allowing classification models to be retrained and updated based on specific measurement tasks and applications. The system can adapt its classification behavior dynamically to different environments, object types, and measurement conditions while maintaining ease of operation through automated model selection and training processes

Inventive Principle:
Principle #15Dynamics

Solution Approach 2:

The system achieves adaptability through parameter changes by adjusting classification parameters and model configurations based on the specific application and environment. Different measurement tasks can utilize customized training data and model parameters, enabling the system to adapt to diverse applications while maintaining user-friendly operation through automated parameter optimization

Inventive Principle:
Principle #35Parameter changes

Data Source

PatentEP3340106B1Method and system for assigning particular classes of interest within measurement data
Publication Date: 2023.02.08 HEXAGON TECH CENT GMBH
  • EP3340106B1 patent drawingFigure 1~2
  • EP3340106B1 patent drawingFigure 3a~3b
  • EP3340106B1 patent drawingFigure 3c~3d

AI summary

The present invention relates to a method and system for surveying and/or metrology for assigning particular classes of interest within measurement data, wherein an assignment - based on a classification model - of at least one measurement object to a first class of interest within the measurement data is processed by a feedback procedure providing feedback data for a training procedure which provides update information for the classification model, wherein the training procedure is based on a machine learning algorithm, e.g. relying on deep learning for supervised learning and/or unsupervised learning.