3D Feature Localization Calibration for Noisy Welding Sensing

Resolve Bottlenecks,
Find Innovative Solutions
Generate Solutions

Solution Overview

Problem

Existing depth-sensing camera systems, such as those used in automated welding processes, face challenges in environments with high noise levels and computational burdens, leading to inaccurate depth determination and unsuitability for applications requiring high accuracy and responsiveness.

Innovation Solution

A calibration method that maps a three-dimensional reference system to the frames of multiple image sensors, allowing for precise depth sensing without the need for specific sensor positioning or focal length measurements, reducing computational load and noise interference.

Engineering Contradictions & Design Principles

VSEngineering Contradiction Analysis

1Measurement precision

If depth-sensing camera systems use triangulation based on signal reflection or stereo-vision with deep learning algorithms, then depth determination capability is improved, but computational burden increases and response time decreases

Engineering Contradiction:
Improvedepth determination accuracyVSAvoidcomputational burden
Core Design Contradiction:
Measurement precisionVSDevice complexity

Solution Approach 1:

The patent applies preliminary action by pre-calibrating the sensing system to establish a mapping between image sensor positions and three-dimensional reference system positions. This calibration data is stored and reused during operation, eliminating the need for complex real-time computations. The system determines depth by looking up pre-computed mapping relationships rather than performing intensive triangulation or stereo-matching algorithms during actual depth sensing operations.

Inventive Principle:
Principle #10Preliminary action

2Measurement precision

If depth-sensing camera systems use triangulation or stereo-vision methods, then depth sensing capability is improved, but performance deteriorates in environments with high noise from external light sources

Engineering Contradiction:
Improvedepth sensing accuracyVSAvoidnoise from external light sources
Core Design Contradiction:
Measurement precisionVSObject-affected harmful factors

Solution Approach 1:

The patent uses an intermediary approach by introducing a calibration layer that maps image sensor coordinates to three-dimensional reference positions. This calibration mapping acts as an intermediary that translates noisy image data into accurate depth information using pre-established relationships. The system determines positions by referencing pre-calibrated mappings rather than directly computing depth from noisy image data, thereby filtering out the harmful effects of external light sources.

Inventive Principle:
Principle #24Intermediary (Mediator)

3Productivity

If automated welding processes use traditional depth-sensing systems, then welding capability is maintained, but accuracy and responsiveness decrease due to computational delays and noise interference

Engineering Contradiction:
Improvewelding speed and accuracyVSAvoiddepth determination accuracy
Core Design Contradiction:
ProductivityVSMeasurement precision

Solution Approach 1:

The patent replaces complex mechanical computation systems with a simplified lookup-based approach. Instead of using heavy computational algorithms for real-time depth sensing during welding operations, the system substitutes these with pre-computed mapping relationships stored in memory. This substitution enables fast, accurate depth determination by simply querying pre-calibrated mappings, thereby improving both welding speed and accuracy without the delays associated with real-time computational processing.

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

Data Source

PatentEP4679371A1Pinpointing object features in 3D space
Publication Date: 2026.01.14 NORSK TITANIUM AS
  • EP4679371A1 patent drawingFigure 1
  • EP4679371A1 patent drawingFigure 2
  • EP4679371A1 patent drawingFigure 3

AI summary

There is provided a method (300) of calibrating a sensing system (100) for determining a position of a feature (236) of an object (10). The method (300) comprises: defining (310) reference positions (181, 182) within a three-dimensional reference system; determining (320, 330) corresponding positions within an image frame (122) of an image sensor (120) and positions within a sensor frame (132) of a further sensor (130); and generating (340) a mapping (150) linking the reference positions and the positions. A position of the feature of the object may be determined based on the mapping. There is also provided a method (400) for monitoring a workpiece (10), the method (400) comprising: obtaining (420, 430) shortwave infrared images (125, 135); determining (440) a corresponding image position of a feature of the workpiece or of an automated welding system using a machine-learned model (190); and determining (450) a position of the feature.