CNN Image Similarity for Hidden-Feature Coating Formula Matching
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing color matching systems for coatings struggle with inefficiencies in analyzing complex mixtures, requiring substantial user intervention and producing inconsistent results due to the inability to account for hidden features beyond measurable spectral and texture features.
Innovation Solution
A computer-implemented method using deep learning techniques, specifically convolutional neural networks, to measure and analyze digital images of coatings, incorporating both measurable and hidden features to improve the accuracy of matching formulas by training an image similarity metric.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Loss of information
If traditional spectrophotometer and image processing algorithms are used to measure and analyze coating properties, then measurable features (spectral data, texture values) can be obtained, but hidden features necessary for accurate color matching cannot be captured
Solution Approach 1:
A convolutional neural network (CNN) is introduced as an intermediary between the measurable image data and the hidden features. The CNN processes the digital images obtained from the spectrophotometer and extracts both measurable and hidden features that are not directly observable through traditional image processing algorithms
Solution Approach 2:
Traditional image processing algorithms are replaced with a deep learning-based convolutional neural network. This substitution enables the system to automatically learn and extract complex features from images without relying on manually designed algorithms, thereby capturing hidden features that traditional methods miss
2Reliability
If manual visual inspection and user intervention are used to evaluate coating matches, then subjective assessment can be performed, but the process becomes time-consuming and produces inconsistent results
Solution Approach 1:
The system performs automatic feature extraction and color match evaluation without requiring manual visual inspection or user intervention. The convolutional neural network autonomously processes the images and determines matching formulas, eliminating the need for human evaluators while maintaining consistent and reliable results
Solution Approach 2:
The system uses training data consisting of triplets (target image, good match image, bad match image) to train the CNN. During training, the network receives feedback on its predictions and adjusts its parameters to minimize errors, thereby learning to consistently distinguish between good and bad color matches
3Productivity
If case-by-case analysis of each unknown target coating is performed using traditional methods, then individual evaluation can be conducted, but the process becomes extremely time-consuming for complex coating mixtures
Solution Approach 1:
The convolutional neural network is trained on a diverse dataset of various coating types and formulations. Once trained, the single CNN model can universally analyze any unknown target coating regardless of its specific composition or complexity, eliminating the need for different analysis procedures for different coating types
Solution Approach 2:
The system performs preliminary training of the CNN on a large dataset of known coating formulas and their corresponding images before actual color matching tasks. This preliminary action enables the system to quickly and efficiently analyze unknown coatings without requiring complex real-time processing during the actual matching process
Data Source
AI summary
Disclosed herein is a method and a device that can measure an unknown target coating; can search, based on the measured data of the target coating, a database for one or more best matching coating formulas, i.e. one or more preliminary matching formulas, within the database; and that can refine the search using an image similarity metric between images of the one or more best matching coating formulas on the one side and images of the target coating on the other side, using deep learning techniques.