Face Collation Using Uncovered Regions Under Masked Conditions
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing face collation technologies struggle to accurately match faces when partially obscured by outfitting members such as masks or sunglasses.
Innovation Solution
A collation device and method that detects feature points in a face region, identifies specific areas not covered by outfitting members, and performs collation using these areas, employing machine-learning models for both full-face and partial-face image processing.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If traditional face collation methods are used, then the collation process is simple, but the collation accuracy deteriorates when the face is partially covered by outfitting members
Solution Approach 1:
The face region is segmented into multiple areas based on coverage detection. The collation process is divided into two paths: one for fully exposed faces and another for partially covered faces. This segmentation allows the system to apply appropriate collation methods to different regions, improving accuracy without requiring a complete redesign of the collation process.
Solution Approach 2:
The collation process dynamically adapts based on the detected coverage state. The system automatically switches between different collation strategies depending on whether outfitting members are detected, making the process flexible and responsive to varying conditions rather than following a fixed procedure.
2Measurement precision
If the entire face region is used for collation, then more feature points are available for matching, but collation accuracy deteriorates when outfitting members cover parts of the face
Solution Approach 1:
The system extracts and identifies the specific areas covered by outfitting members, then excludes these regions from the collation process. By removing the problematic covered areas from consideration, the system prevents these regions from degrading the overall collation accuracy while still utilizing all available uncovered feature points.
Solution Approach 2:
Different quality standards are applied to different regions of the face. Uncovered regions are used for collation with full weight, while covered regions are excluded or given minimal weight. This local differentiation ensures that high-quality uncovered feature points drive the collation result rather than being diluted by low-quality covered regions.
3Adaptability or versatility
If a fixed collation method is used, then the process is straightforward, but adaptability deteriorates when facing different coverage scenarios
Solution Approach 1:
The collation system is designed to perform multiple functions: it can handle both fully exposed faces and partially covered faces using a unified framework. The coverage detection module and adaptive collation logic enable the same system to adapt to various scenarios without requiring separate dedicated systems for each case.
Solution Approach 2:
The system incorporates feedback through coverage detection that informs the collation process. By continuously monitoring the coverage state and adjusting the collation strategy accordingly, the system creates a closed-loop adaptive process that responds to actual conditions rather than following a predetermined fixed path.
Data Source
AI summary
A feature point is detected from a face region of a person reflected in an image. A specific area in which the person's face is not covered with an outfitting member is determined in the face region. Collation is performed using the feature point in the specific area.