Surface Normal Imaging for Color-Invariant Defect Detection in Leather
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current methods for automatic defect detection in animal or leather skins are not highly reliable and user-friendly, relying on brightness or color values, which can be influenced by color differences and are not effective in accurately identifying imperfections.
Innovation Solution
The method evaluates the orientation or alignment of surface units using 3D information, such as surface normals, rather than brightness values, allowing for more accurate and user-friendly automatic defect detection, independent of color differences.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If brightness or color values are used for defect detection, then the detection process is simple, but the detection reliability is low due to influence from color differences
Solution Approach 1:
The patent transitions from 2D brightness/color value analysis to 3D surface orientation analysis by calculating surface normals from multiple images taken under different lighting conditions. This dimensional change enables reliable defect detection independent of color variations, as surface geometry remains invariant to color changes.
Solution Approach 2:
The patent changes the detection parameter from brightness/color values to surface orientation parameters (surface normals). By evaluating the orientation of surface units rather than their reflectivity, the system achieves color-invariant defect detection, fundamentally changing what parameter is being measured to solve the reliability issue.
2Measurement precision
If 3D surface orientation information is used for defect detection, then detection accuracy improves, but the complexity of the detection system increases
Solution Approach 1:
The patent captures 3D surface orientation information by taking multiple 2D images under different lighting angles and computing surface normals. This approach achieves precise 3D measurement precision while using relatively simple 2D imaging devices, effectively adding a dimension of information without proportionally increasing device complexity.
Solution Approach 2:
The patent uses surface normals as an intermediary representation that bridges the gap between simple 2D image capture and complex 3D defect characterization. By computing and analyzing surface normals rather than directly measuring complex 3D geometry, the system achieves high precision with manageable processing complexity.
3Reliability
If multiple images under different lighting situations are used to calculate surface normals, then defect detection becomes independent of color, but the processing time increases
Solution Approach 1:
The patent performs preliminary computation of surface normals from multiple images before the actual defect detection and classification processes. By pre-calculating the orientation information, the system enables rapid defect analysis in subsequent steps, as the color-invariant 3D surface characteristics are already extracted and ready for evaluation.
Data Source
Figure 1
Figure 2
Figure 3a~3b
AI summary
Described and illustrated is a method for detecting defects in flexible bodies (11), particularly animal hides, comprising the steps of: - providing one or more, in particular at least three, and further in particular at least five, light source(s) (16), preferably point light source(s), for illuminating at least one sub-area of the body (11); - taking multiple images of at least one sub-area of the body (11) by means of a recording device (15) comprising, in particular, at least one camera (15a, b, c); - each of the multiple images being taken under a different illumination situation generated by the light source(s) (16), in particular such that a different one of the multiple light sources (16) is activated for each image; - calculating orientation information, in particular the normal vector, of micro-areas (30), for example pixels, assigned to the body (11), based on the multiple images.- Analysis of alignment information to detect body defects (11).,