Effect Pigment Identification Through Pixel-Wise Neural Segmentation
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing methods for identifying and matching effect pigments in coatings, such as automotive coatings, are inefficient and often require time-consuming, subjective processes that fail to accurately account for interactions between different types of effect pigments, leading to inconsistent results.
Innovation Solution
A computer-implemented method using semantic segmentation and neural networks to analyze digital images of coatings, annotating pixels with pigment labels, and determining pigment statistics to facilitate efficient database searches for matching formulas, incorporating both pixel-wise and sparkle point classification techniques.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If manual visual inspection and case-by-case analysis are used to identify effect pigments, then measurement precision may be maintained through human expertise, but productivity is severely reduced due to time-consuming analysis of each target coating
Solution Approach 1:
The patent replaces manual visual inspection and human expert analysis with an automated image processing system using cameras, spectrometers, and computational algorithms. The system captures images of the target coating, processes them through image analysis algorithms, and automatically identifies effect pigments and their interactions, eliminating the need for time-consuming manual case-by-case analysis while maintaining identification accuracy through objective computational methods
Solution Approach 2:
The patent transforms the analysis approach by changing from qualitative human visual assessment to quantitative image parameter analysis. By capturing digital images and extracting parameters such as color values, texture features, sparkle intensity, and hue distribution, the system enables automated comparison against reference databases, significantly improving analysis speed while maintaining precision through objective parameter measurement
2Measurement precision
If traditional image processing algorithms are used to derive texture values, then device complexity is kept simple, but measurement precision is insufficient for identifying specific effect pigments and their interactions
Solution Approach 1:
The patent applies image segmentation techniques to divide the coating image into distinct regions corresponding to different effect pigment types. By segmenting the image based on color, texture, and optical properties, the system can identify and analyze specific pigment regions separately, enabling precise identification of multiple effect pigments and their interactions while managing computational complexity through localized analysis
Solution Approach 2:
The patent enhances traditional 2D image analysis by incorporating spectral dimensionality through photospectrometer data. By adding the spectral dimension to the analysis, the system can distinguish between different effect pigment types based on their unique spectral signatures, significantly improving measurement precision for identifying specific pigments and their interactions without excessive complexity increase
3Ease of manufacture
If training data is generated using measurements of colors including only one type of effect pigment, then ease of manufacture for training data is improved, but reliability of the neural network is reduced due to inability to learn interactions between different types of effect pigments
Solution Approach 1:
The patent creates composite training data by combining measurements from multiple types of effect pigments in various combinations. Instead of training on single-pigment colors only, the system generates training datasets that include complex multi-pigment formulations, enabling the neural network to learn the interactions between different effect pigment types. This approach maintains ease of data generation by using the same measurement instrumentation while significantly improving classification reliability through exposure to diverse pigment combinations
Data Source
AI summary
Described herein is a computer-implemented method. The method includes: providing digital images and respective formulas for coating compositions with known pigments and/or pigment classes associated with the respective digital images, classifying, using an image annotation tool, for each digital image, each pixel, by visually reviewing the respective digital image pixel-wise, providing, for each digital image, an associated pixel-wise annotated image, training a first neural network with the provided digital images as input and the associated pixel-wise annotated images as output, making the trained first neural network available for applying the trained first neural network to at least one unknown input image of a target coating and for assigning a pigment label and/or a pigment class label to each pixel in the at least one unknown input image, and determining and/or outputting, for each unknown input image, a statistic of corresponding identified pigments and/or pigment classes, respectively.