3D Model Classification for Sasang Constitution Diagnosis
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current methods for diagnosing sasang constitution using image information face challenges in correcting color distortions and geometric distortions caused by varying lighting conditions and camera vibrations, leading to potential misdiagnosis due to inconsistent color references and three-dimensional face structure analysis limitations.
Innovation Solution
A three-dimensional model classification method that corrects color values in frontal and profile images using a reference color table, generates a three-dimensional geometric model by extracting feature point information, and classifies constitutions by matching the model with standardized reference models to minimize spatial displacements.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If color correction is performed using reference color tables in image diagnosis, then color accuracy is improved, but the complexity of image processing increases
Solution Approach 1:
The patent introduces a reference color table as an intermediary element that mediates between the captured image and the desired color correction. The reference color table serves as a mediator to transform color values from the distorted image color space to the standardized reference color space, thereby achieving accurate color correction without requiring complex direct correction algorithms.
Solution Approach 2:
The patent applies parameter changes by transforming color values through mathematical operations. Specifically, the color values in the captured image are multiplied by correction matrices derived from the reference color table, changing the color parameters (RGB values) to achieve accurate color representation while maintaining a systematic and manageable processing approach.
2Measurement precision
If three-dimensional geometric model is generated from multiple images, then constitution classification accuracy is improved, but the difficulty of extracting accurate geometric information increases
Solution Approach 1:
The patent replaces complex mechanical/visual processing of three-dimensional geometry with a simplified feature-based approach. Instead of directly processing complex 3D geometric information from multiple images, the system extracts key feature points (eyes, nose, mouth, ears) and uses their positional relationships to represent the face structure, thereby reducing the difficulty of geometric information extraction while maintaining classification accuracy.
Solution Approach 2:
The patent segments the face into distinct feature points and regions (eyes, nose, mouth, ears, forehead, chin). By dividing the complex three-dimensional face structure into discrete, identifiable features, the system simplifies the extraction process and enables accurate constitution classification based on the relative positions and characteristics of these segmented features.
3Measurement precision
If feature point information is extracted from corrected images, then model comparison precision is improved, but the time required for image processing increases
Solution Approach 1:
The patent applies preliminary action by performing color correction using the reference color table before extracting feature point information. This preliminary color correction ensures that the subsequent feature extraction and model comparison are performed on accurately colored images, improving precision. The correction matrices are pre-calculated and stored, allowing for efficient application during actual processing.
Solution Approach 2:
The patent uses partial action by extracting only the necessary feature point information (eyes, nose, mouth, ears, and their positions) rather than processing every pixel and detail of the image. This selective extraction approach maintains sufficient precision for constitution classification while significantly reducing the time required compared to comprehensive image analysis.
Data Source
AI summary
Provided is a three-dimensional model classification method of classifying constitutions. The method includes correcting color values of a frontal image and one or more profile images to allow a color value of a reference color table in the images to equal a predetermined reference color value, through obtaining the frontal image and one or more profile images of a subject including the reference color table by a camera, the reference color table including one or more sub color regions, generating a three-dimensional geometric model of the subject by extracting feature point information from the frontal image and the profile image, matching the corresponding feature point information to extract spatial depth information, after removing the reference color table region from the frontal image and the profile image, and classifying a group of the three-dimensional geometric model of the subject by selecting a reference three-dimensional geometric model having a smallest sum of spatial displacements from the three-dimensional geometric model of the subject from a plurality of reference three-dimensional geometric models stored in the database and setting the group which the selected reference three-dimensional geometric model represents as the group where the three-dimensional geometric model of the subject belongs.