Intrasensor Uniformity Correction for Optical Web Defect Detection
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Conventional visual inspection systems for moving webs suffer from intra-device detection non-uniformity, where identical web anomalies are classified differently due to variations in optical arrangement, illumination, and material characteristics across the field of view, leading to inconsistent defect detection.
Innovation Solution
A flexible intra-device image correction technique that allows selection and adjustment of pixel normalization algorithms, such as gain-based and offset-based corrections, to normalize image data uniformly across the field of view, ensuring consistent defect detection regardless of anomaly location.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If conventional image capture devices are used without correction, then the system is simple and fast, but intra-device non-uniformity causes identical web anomalies to be classified differently across the field of view
Solution Approach 1:
The patent applies preliminary action by pre-calculating and storing normalization values for each pixel location in the image capture device before actual defect detection. These normalization values compensate for intra-device non-uniformity characteristics. During operation, the pre-computed normalization values are applied to raw image data to correct systematic variations across the field of view, ensuring consistent defect classification without adding real-time computational complexity.
Solution Approach 2:
The patent changes the parameter of pixel response normalization by applying location-specific normalization factors to each pixel in the image data. Instead of uniform processing, each pixel position (x, y) is normalized using its specific normalization value that accounts for optical variations, illumination gradients, and material characteristics at that particular location in the field of view.
2Adaptability or versatility
If multiple normalization algorithms are combined with adjustable coefficients, then flexibility and accuracy improve, but the complexity of selecting and adjusting normalization parameters increases
Solution Approach 1:
The patent enables parameter changes by allowing dynamic adjustment of coefficients (e.g., a1, a2, a3) that control the contribution of different normalization algorithms (such as offset-based, gain-based, and polynomial corrections). Users can modify these parameters to optimize performance for specific web materials, illumination conditions, or defect types, providing adaptability without requiring changes to the underlying normalization algorithms.
Solution Approach 2:
The patent achieves universality by designing a multi-algorithm normalization system where multiple normalization approaches (offset correction, gain correction, polynomial fitting) are integrated into a single unified framework. The system can selectively apply different algorithms or combinations thereof depending on the specific application requirements, making it versatile across various web materials and inspection scenarios.
3Measurement precision
If location-specific normalization values are applied to each pixel, then uniform defect classification across the field of view is achieved, but computational processing time increases
Solution Approach 1:
The patent resolves the time-constraint contradiction by performing the computationally intensive normalization value calculations in advance, before actual defect detection begins. The normalization values for each pixel location are pre-computed based on calibration data or historical measurements, stored in memory, and then rapidly applied during production inspection. This eliminates real-time computational delays while maintaining precise location-specific normalization.
Solution Approach 2:
The patent substitutes complex real-time computational mechanics with simpler data retrieval and application operations. Instead of calculating normalization values on-the-fly during defect detection, the system replaces this with memory lookup and straightforward multiplication operations using pre-stored normalization factors, significantly reducing processing time while preserving measurement precision.
Data Source
AI summary
Techniques are described in which an image capture device captures image data from web material. The image data comprises pixel values for the cross-web field of view of the image capture device. An analysis computer includes a computer-readable medium that stores parameters for a plurality of different normalization algorithms to normalize a cross-web background signal for the image capture device to a common desired value. The computer-readable medium further stores coefficients specifying a weighting for each of the plurality of normalization algorithms. The analysis computer computes a normalized value for each of the pixels of the image data as a weighted summation of results from application of at least two of the pixel normalization algorithms using the stored parameters.