Welding Mark Classification Using Segmentation to Avoid Misjudgment
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing battery production processes face challenges in accurately inspecting welding marks due to background interference and reliance on gray level thresholds, leading to misjudgment and increased overkill and underkill rates.
Innovation Solution
A welding mark inspection method and apparatus that uses a classification model to analyze characteristics such as shape, color, and area without setting a gray level threshold, segmenting the welding mark from the image to reduce interference, and performing secondary inspections on ambiguous cases to improve accuracy.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Ease of operation
If a gray level threshold is used to judge welding mark quality, then the inspection process is simple, but misjudgment occurs due to influence from shooting position, shooting device, or other factors on image gray level
Solution Approach 1:
The patent changes the inspection parameter from gray level (which is sensitive to shooting conditions) to multiple characteristics including shape, area, color, and texture. This parameter transformation allows the system to maintain inspection simplicity while significantly improving accuracy by selecting features that are invariant to shooting position and device variations.
Solution Approach 2:
The patent replaces the simple threshold-based mechanical judgment system with an intelligent classification model (machine learning-based system). This substitution enables the system to automatically learn and adapt to different welding mark characteristics without manual threshold adjustment, resolving the contradiction between operational simplicity and measurement precision.
2Reliability
If the entire picture is processed for welding mark inspection, then comprehensive inspection is achieved, but processing time increases and efficiency decreases
Solution Approach 1:
The patent segments the inspection process into distinct stages: first locating the welding mark region in the picture, then extracting specific characteristics (shape, area, color, texture) only from that region, and finally classifying the welding mark quality. This segmentation allows comprehensive inspection of welding marks while avoiding unnecessary processing of the entire picture, thus improving efficiency.
Solution Approach 2:
The patent extracts only the relevant welding mark region and its key characteristics from the entire picture for classification. By taking out and focusing only on the essential features (shape, area, color, texture) of the welding mark rather than processing all picture data, the system achieves both comprehensive inspection and high processing efficiency.
3Loss of information
If background regions are included in welding mark analysis, then complete picture information is available, but classification accuracy decreases due to interference from non-welding mark regions
Solution Approach 1:
The patent extracts and isolates the welding mark region from the background by locating the welding mark position first, then analyzing only the characteristics within that specific region. This extraction process removes interfering background information while preserving all essential welding mark information, thereby improving classification accuracy without losing relevant data.
Solution Approach 2:
The patent segments the picture into welding mark region and background region, then processes only the welding mark region for quality classification. This segmentation strategy maintains the completeness of picture information for location purposes while eliminating background interference from the classification analysis, resolving the contradiction between information completeness and classification accuracy.
Data Source
AI summary
A welding mark inspection method includes acquiring a picture including a welding mark, and obtaining a classification type of the welding mark based on the picture and a welding mark classification model. The welding mark classification model is configured to classify the welding mark based on characteristics of the welding mark.