3D Feature Localization Calibration for Noisy Welding Sensing
Find Innovative SolutionsGenerate 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
Engineering 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
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.
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
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.
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
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.
Data Source
Figure 1
Figure 2
Figure 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.