Battery Cell Surface Defect Detection With Image Preprocessing
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing defect detection methods in battery cell surfaces are inefficient and inaccurate, leading to low production yield due to issues with feature extraction and reliance on imaging quality, which affects the recognizability of defects like indentation defects.
Innovation Solution
A method and apparatus using computer vision and deep learning technologies to preprocess images of battery cell surfaces, extracting and segmenting cell surface areas, and utilizing a defect detection neural network model to enhance feature prominence and accuracy.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If traditional defect detection methods are used on battery cell surfaces, then the detection process is simple, but the detection accuracy and efficiency are low due to poor feature extraction and reliance on imaging quality
Solution Approach 1:
The patent applies preliminary action by performing image preprocessing operations (exposure adjustment, noise filtering, contrast enhancement) before the actual defect detection. This prepares the images in advance to optimize feature visibility, allowing the detection model to work with pre-optimized data and achieve higher accuracy without increasing the complexity of the core detection algorithm.
Solution Approach 2:
The patent segments the battery cell surface into multiple regions of interest (ROIs) based on predefined geometric shapes or detected contours. This segmentation allows the detection system to focus computational resources on specific areas where defects are more likely to occur, improving detection accuracy while managing system complexity through localized analysis rather than processing the entire surface uniformly.
2Measurement precision
If the entire cell surface is processed as a single image, then the processing is straightforward, but small defects are difficult to detect due to their small proportion of the total surface area
Solution Approach 1:
The patent divides the cell surface image into multiple smaller regions or patches, allowing small defects to occupy a larger relative area within their local region. This segmentation makes subtle defect features more prominent and easier to detect, while the modular approach enables parallel processing of multiple regions, reducing overall detection time compared to analyzing the entire large image sequentially.
Solution Approach 2:
The patent applies local quality by using different processing parameters or detection thresholds for different regions of the cell surface. Regions with higher defect probability or different surface characteristics can be analyzed with optimized settings, improving small defect detection accuracy in critical areas without uniformly increasing processing time across the entire surface.
3Measurement precision
If the exposure of the cell surface area is not adjusted, then the imaging process is simple, but defect features are not prominent enough for accurate detection
Solution Approach 1:
The patent changes the exposure parameter of the cell surface area to optimize defect feature visibility. By adjusting exposure levels, the imaging system enhances the contrast between defects and the surrounding surface, making defect features more prominent. This parameter adjustment is applied automatically through image processing algorithms, maintaining ease of manufacture while significantly improving detection precision.
4Measurement precision
If conventional machine learning models are used for defect detection, then the model structure is simple, but the detection accuracy is insufficient for complex defect patterns
Solution Approach 1:
The patent replaces conventional machine learning models with a deep learning neural network model. This substitution enables the system to automatically learn complex defect patterns and features from training data, achieving superior detection accuracy for diverse and complex defect types. The deep learning model, while more complex in structure, provides significantly improved performance that justifies the increased model complexity.
Data Source
AI summary
A method and apparatus for detecting a defect on a surface of a cell, including: obtaining an initial image of the surface of the cell by using an image acquisition unit; preprocessing the initial image to obtain at least one image to be detected of the surface of the cell; and inputting the at least one image to be detected into a defect detection neural network model, and obtaining a detection result outputted by the defect detection neural network model, the detection result being used to indicate whether there is a defect on the surface of the cell is described. According to technical solutions, features of the defect in the image may be made prominent, so as to facilitate subsequent detection of the defect detection neural network model; and more features may be extracted from the image, thereby improving the efficiency and accuracy of defect detection.