Surface Property Estimation System Using Machine Learning
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing methods for estimating surface properties of objects and extracting unique identifiers from images are not effectively integrated, making it difficult to associate surface properties with individual identifiers for efficient product management and quality control.
Innovation Solution
A surface property estimation system that includes image acquisition, estimation using machine learning models, feature extraction, and registration of surface properties and identifiers in a storage medium, allowing for the capture of surface images, estimation of properties, extraction of unique features, and their association for storage and matching.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If a three-dimensional measuring sensor is used to measure surface property, then measurement accuracy is improved, but cost and measurement time increase
Solution Approach 1:
The patent replaces the mechanical three-dimensional measuring sensor system with an optical imaging system combined with machine learning algorithms. Instead of using dedicated measurement hardware that requires skilled operation, the system uses standard imaging devices to capture surface images and processes them through trained estimation models, thereby substituting mechanical measurement with optical capture and computational analysis.
Solution Approach 2:
The patent creates a virtual model (estimation model) through machine learning that replicates the measurement capability of expensive three-dimensional sensors. By training the model with pairs of images and actual surface property measurements, the system creates a digital copy of the measurement function that can be executed on standard imaging hardware, eliminating the need for costly dedicated measurement devices.
2Adaptability or versatility
If individual product management methods (manufacturing numbers, barcodes, RFID tags) are applied to each product, then product identification capability is improved, but cost increases proportionally with product volume
Solution Approach 1:
The patent enables products to self-identify through their own surface characteristics rather than requiring external tags or labels. By extracting unique feature amounts directly from the product surface images, the system allows each product to serve its own identification function, eliminating the need for separate identification components and reducing per-unit costs.
Solution Approach 2:
The patent makes the imaging device serve multiple functions: it both captures images for surface property estimation and extracts unique identifiers for product identification. This multi-functional approach eliminates the need for separate identification systems, reducing overall system cost while maintaining identification capability across all products.
3Ease of manufacture
If surface property estimation and identifier extraction are performed separately, then each function can be optimized independently, but integration and association of data become difficult
Solution Approach 1:
The patent merges the surface property estimation function and the identifier extraction function into a single integrated processing system. Both functions operate on the same input image data and their results are associated through the common image identifier, eliminating the need for separate processing pipelines and simplifying data correlation while maintaining functional optimization.
Data Source
AI summary
A surface property estimation system includes an image acquisition means for acquiring an image of a surface of an object, an estimation means for estimating a surface property from the acquired image by using an estimation model obtained through machine learning with use of an image of a surface of an object and a surface property shown by the image as training data, an extraction means for extracting, from the acquired image, a feature amount unique to the image, and a registration means for storing the estimated surface property and the extracted feature amount in a storage means in association with each other.