Weld Bead Image Recognition Using Context-Aware Segmentation
Find Innovative SolutionsGenerate 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
Engineering 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
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
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
2Measurement precision
If weld bead positioning is performed without contextual information, then the process is simple, but the accuracy of defect detection is insufficient
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
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
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
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
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
Data Source
Figure 1
Figure 2(a)~3
Figure 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.