3D Image Match Probability via Zebra Shading and CNN
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing methods for acquiring probabilities of match and mismatch between three-dimensional CAD images and actual images, as well as between simulated and actual images, lack accuracy, necessitating a more precise approach for comparing these images.
Innovation Solution
A probability acquisition apparatus that includes a first image acquisition unit for obtaining a 3D image from data, a second image acquisition unit for obtaining a 3D simulated image, a zebra image acquisition unit for processing both images into zebra images with identical angles, scales, and sizes, and a machine learning model to determine probabilities of match and mismatch between selected image regions after noise removal and feature extraction.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If conventional image processing methods are used to compare 3D-CAD images and actual images, then the comparison can be performed, but the accuracy of match/mismatch probabilities is insufficient
Solution Approach 1:
The patent introduces zebra images as an intermediary representation between the original 3D images and the machine learning model. These zebra images encode depth and surface information through alternating light-dark patterns, serving as a mediator that enhances the discriminative features for probability calculation while maintaining system modularity
Solution Approach 2:
The patent replaces conventional image processing algorithms with a machine learning model (neural network) that automatically learns optimal feature representations. This substitution enables the system to achieve higher accuracy in probability calculation without manually designing complex processing pipelines
2Reliability
If the entire image is processed to ensure comprehensive comparison, then completeness is improved, but the operational load increases
Solution Approach 1:
The patent divides the image processing into distinct segments: zebra image generation, region of interest selection, and probability calculation. This segmentation allows the system to process only relevant regions (hands, faces, feet) rather than entire images, reducing computational load while maintaining reliability for critical comparison areas
Solution Approach 2:
The patent applies partial processing by focusing computational resources on selecting and processing only the most relevant image regions (hands, faces, feet) rather than processing the entire image. This partial action approach maintains reliability for key identification areas while significantly reducing overall operational load
Data Source
AI summary
To provide a probability acquisition apparatus and the like that can provide accurate probabilities of a match and a mismatch between two three-dimensional images of a target object, a probability acquisition apparatus performs zebra shading processing on a first image that is a three-dimensional image of an automotive clay model and a second image that is a 3D-CAD image to obtain a first zebra image and a second zebra image, obtains a first selected image from the first zebra image with noise removed and a second selected image from the second zebra image with features extracted, and inputs a superimposition image of the first selected image and the second selected image to a CNN 15b to thereby obtain a probability of a match between the first selected image and the second selected image.