Weld Bead Image Recognition Using Context-Aware Segmentation

Resolve Bottlenecks,
Find Innovative Solutions
Generate Solutions

Solution Overview

Problem

Existing weld bead positioning methods in battery manufacturing are inaccurate due to variations in temperature and environmental factors, leading to defects such as pinholes and false welding, which affect the safety performance of batteries.

Innovation Solution

An image recognition method involving coarse-to-fine segmentation, feature extraction, and context representation using convolutional neural networks to enhance pixel correlations, followed by parameter fitting of contour points for improved weld bead recognition.

Engineering Contradictions & Design Principles

VSEngineering Contradiction Analysis

1Reliability

If traditional welding methods are used to seal battery filling ports, then the sealing process can be completed, but temperature and environmental variations cause welding defects such as pinholes, burst points, and false welding

Engineering Contradiction:
Improvesealing reliabilityVSAvoidtemperature and environmental variations
Core Design Contradiction:
ReliabilityVSObject-affected harmful factors

Solution Approach 1:

The patent applies preliminary action by performing pre-welding position detection and contour extraction before the actual welding process. The system detects the welding position, extracts the weld bead contour, and stores this information for later reference during welding, allowing the welding process to proceed with predetermined guidance to avoid defects caused by temperature and environmental variations

Inventive Principle:
Principle #10Preliminary action

Solution Approach 2:

The patent implements feedback by using the detected welding position and extracted contour information to guide the welding process. The system compares actual welding progress with the pre-detected contours and adjusts welding parameters in real-time, providing feedback control to maintain welding quality despite temperature and environmental fluctuations

Inventive Principle:
Principle #23Feedback

2Measurement precision

If weld bead positioning is performed without contextual information, then the process is simple, but the accuracy of defect detection is insufficient

Engineering Contradiction:
Improveweld bead positioning accuracyVSAvoidimage processing complexity
Core Design Contradiction:
Measurement precisionVSDevice complexity

Solution Approach 1:

The patent applies segmentation by dividing the image processing into distinct stages: initial segmentation to obtain first recognition results, feature extraction to obtain region representations, and context representation generation to capture pixel correlations. This segmented approach improves positioning accuracy by processing information in manageable stages while maintaining overall system organization

Inventive Principle:
Principle #1Segmentation

Solution Approach 2:

The patent applies dimensionality change by transitioning from basic pixel-level analysis to region-level feature extraction, and finally to context-level pixel correlation analysis. This multi-dimensional processing approach enriches the positioning information by adding contextual dimensions while managing complexity through hierarchical organization

Inventive Principle:
Principle #17Another dimension (Dimensionality change)

3Productivity

If defect detection is performed on weld beads with variations in temperature and environment, then comprehensive inspection can be conducted, but the accuracy is reduced due to defects like pinholes and false welding

Engineering Contradiction:
Improvedefect detection capabilityVSAvoiddefect detection accuracy
Core Design Contradiction:
ProductivityVSMeasurement precision

Solution Approach 1:

The patent applies preliminary action by extracting and storing the weld bead contour information before defect detection begins. This pre-extracted contour serves as a reference framework that maintains its geometric accuracy even when temperature and environmental factors cause welding defects, enabling reliable defect detection against a stable geometric baseline

Inventive Principle:
Principle #10Preliminary action

Solution Approach 2:

The patent uses the extracted contour information as an intermediary reference that bridges the gap between the original weld bead geometry and the defective final state. This intermediary contour allows the system to detect and measure defects such as pinholes and false welding by comparing actual weld features against the expected geometric pattern

Inventive Principle:
Principle #24Intermediary (Mediator)

Data Source

PatentEP4209995B1Image recognition method and apparatus, and computer-readable storage medium
Publication Date: 2026.01.07 CONTEMPORARY AMPEREX TECHNOLOGY (HONG KONG) LIMITED
  • EP4209995B1 patent drawingFigure 1
  • EP4209995B1 patent drawingFigure 2(a)~3
  • EP4209995B1 patent drawingFigure 4

AI summary

The embodiments of the present application relate to the technical field of image processing, and provide an image recognition method and apparatus, and a computer-readable storage medium. The image recognition method includes: acquiring a target image, where the target image includes a weld bead region; performing initial segmentation on the target image, to obtain a first recognition result, where the first recognition result includes first recognition information for the weld bead region in the target image; performing feature extraction on the target image, to obtain a region representation; obtaining a context representation based on the first recognition result and the region representation, where the context representation is used for representing a correlation between each pixel and remaining pixels in the target image; and obtaining a second recognition result based on the context representation, where the second recognition result includes second recognition information for the weld bead region in the target image. In the embodiments of the present application, the accuracy of image recognition is improved.