Stochastic Sub-Image Analysis for Low-Contrast Defect Detection
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing methods for automatic defect detection in animal skins are unreliable, particularly for flaws with low contrast such as scars and insect bites, as they rely on manual marking or unsuitable automatic evaluation techniques.
Innovation Solution
The method involves dividing the animal skin into small sub-images and performing higher-order stochastic analysis to account for spatial relationships between pixels, using occurrence probability matrices to identify and classify defects, eliminating the need for manual marking and improving detection reliability.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If discontinuity criteria are used to detect flaws automatically, then edge detection is reliable, but low-contrast flaws such as scars and insect bites are not reliably identified
Solution Approach 1:
The animal skin image is divided into multiple small sub-images or windows, each of which is then analyzed separately using higher-order stochastic analysis. This segmentation allows the system to detect local variations and low-contrast flaws that would be missed in a global analysis of the entire skin surface.
Solution Approach 2:
The patent transitions from first-order stochastic analysis (basic histogram analysis) to higher-order stochastic analysis (second, third, or fourth order), which examines spatial relationships between pixels. This parameter change enables the system to detect subtle contrast variations and neighborhood patterns characteristic of low-contrast flaws like scars and insect bites.
2Ease of manufacture
If homogeneity criteria with threshold values are used, then processing is simplified, but the method is unreliable for animal skins with variable texture
Solution Approach 1:
Instead of using fixed threshold values, the patent employs stochastic analysis that calculates probability distributions and higher-order moments from the pixel data itself. This adaptive approach automatically adjusts to the variable texture and lighting conditions of animal skins, maintaining reliability without requiring manual threshold calibration.
Solution Approach 2:
The system uses the actual pixel data to generate statistical models and probability distributions that feedback into the detection process. By continuously analyzing the stochastic properties of the image data, the system adapts to the specific characteristics of each animal skin, improving reliability while maintaining automated processing.
3Reliability
If manual marking methods are used, then detection reliability is maintained, but the process is cumbersome and time-consuming
Solution Approach 1:
The patent replaces manual visual inspection and marking with automated image processing using higher-order stochastic analysis. The system automatically analyzes sub-images, calculates probability distributions, and identifies flaws without human intervention, thereby maintaining detection reliability while significantly improving processing efficiency and productivity.
Data Source
AI summary
The invention relates, among other things, to a method for automatic defect detection in flexible bodies (12), in particular in animal hides, comprising the steps of: - creating an image (15, 29) of at least a partial area of the body (12) by means of an image recording device (13), in particular comprising at least one CCD camera (14), - dividing the image (15, 29) into sub-images (33), - automatic analysis of the sub-images by evaluating the grayscale values or color values of a sub-image (33) taking into account the proximity relationships of the individual pixels of the sub-image (33) to each other.