Object Annotation Correction in Image Data Analysis

Resolve Bottlenecks,
Find Innovative Solutions
Generate Solutions

Solution Overview

Problem

Existing methods for quality control of object markings in images do not allow for automated quantification of deviations from ideal positions or sizes, preventing automatic correction of object markings and analysis of error structures in image data sets.

Innovation Solution

A method and analysis unit that read image data to detect and correct object markings by determining distances between marking sections and object parts, using threshold values to assess correctness and output signals for incorrect definitions, allowing for automatic adjustment of object markings to ideal positions and sizes.

Engineering Contradictions & Design Principles

VSEngineering Contradiction Analysis

1Measurement precision

If manual quality control of object markings is performed, then accuracy of marking evaluation can be maintained, but productivity and automation level are reduced

Engineering Contradiction:
Improveaccuracy of marking evaluationVSAvoidproductivity
Core Design Contradiction:
Measurement precisionVSProductivity

Solution Approach 1:

The system performs self-validation by automatically detecting objects and their parts, then verifying whether object markers correctly encompass them. The analysis unit independently evaluates marking accuracy without requiring manual intervention, achieving both high accuracy and automated operation.

Inventive Principle:
Principle #25Self-service

Solution Approach 2:

The system generates feedback signals indicating whether object markers are correctly defined based on automated detection results. This feedback mechanism enables continuous quality control and automatic correction of marking errors, maintaining accuracy while enabling high-speed automated processing.

Inventive Principle:
Principle #23Feedback

2Productivity

If automated object marking correction is implemented, then productivity and efficiency are improved, but device complexity increases

Engineering Contradiction:
Improveefficiency of quality controlVSAvoidcomplexity of analysis system
Core Design Contradiction:
ProductivityVSDevice complexity

Solution Approach 1:

The analysis method is divided into distinct functional modules: object detection, object part detection, distance calculation, threshold comparison, and signal generation. This segmentation allows each module to be independently optimized and maintained, reducing overall system complexity while enabling automated correction.

Inventive Principle:
Principle #1Segmentation

Solution Approach 2:

The analysis unit acts as an intermediary between image data and quality control decisions. It receives image data, performs automated analysis, and outputs correction signals, simplifying the overall system architecture by centralizing the complex analysis functions in a dedicated component.

Inventive Principle:
Principle #24Intermediary (Mediator)

3Measurement precision

If detailed distance measurements and threshold comparisons are performed, then measurement precision of marking deviations is improved, but loss of time in processing increases

Engineering Contradiction:
Improveprecision of deviation quantificationVSAvoidprocessing time
Core Design Contradiction:
Measurement precisionVSLoss of time

Solution Approach 1:

The system performs distance measurements and threshold comparisons only for critical object parts that require precise marking validation. By focusing computational resources on essential measurements rather than exhaustive analysis of all image elements, the system achieves high precision for marking deviations while minimizing processing time.

Inventive Principle:
Principle #16Partial or excessive action

Data Source

PatentEP3529745B1Method for object annotation in images
Publication Date: 2024.10.09 BAYERISCHE MOTOREN WERKE AG
  • EP3529745B1 patent drawingFigure 1
  • EP3529745B1 patent drawingFigure 2
  • EP3529745B1 patent drawingFigure 3

AI summary

Disclosed is an analysis method for checking at least one object label in image data, wherein: the analysis method reads data of at least one image from a memory; the image displays a pre-defined object; a pre-defined object portion of the pre-defined object is specified and an object label that encloses the pre-defined object is defined for said pre-defined object; the analysis method detects the pre-defined object without an object portion in the data of the image and if the analysis method does not detect the pre-defined object without an object portion in the image, issues a first signal indicating that the object label is incorrectly defined.