Partial Face Recognition via Boarding Condition Learning
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing face recognition technologies struggle to accurately identify users when a portion of their face is covered, leading to incorrect recognition and inconvenience in vehicle access systems.
Innovation Solution
A face recognition apparatus and method that learns a boarding condition, including location and time, to reduce the number of feature points required for recognition, allowing accurate identification using partial-face images when the current boarding condition matches the learned condition, even if the face is partially obscured.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If full-face image is required for face recognition, then recognition accuracy is improved, but recognition fails when face is partially covered
Solution Approach 1:
The patent applies partial action by extracting and matching only the visible feature points from the partial-face image rather than requiring all feature points. The system identifies which feature points are visible and performs recognition based on those subset points, enabling successful recognition even when part of the face is covered.
Solution Approach 2:
The patent changes the parameter of feature point quantity from fixed (all points required) to variable (only visible points used). The system dynamically adjusts the number and type of feature points based on what is visible in the partial-face image, allowing flexible adaptation to different coverage conditions while maintaining recognition accuracy.
2Measurement precision
If boarding condition learning is implemented, then recognition accuracy under partial coverage is improved, but system complexity increases
Solution Approach 1:
The patent applies preliminary action by pre-learning and storing the boarding conditions (location and time patterns) of registered users before actual recognition occurs. When a passenger boards, the system checks whether current conditions match the pre-learned patterns, and only then proceeds to perform face recognition with reduced feature points, thereby improving accuracy without requiring complex real-time analysis.
3Speed
If number of feature points is reduced, then recognition speed is improved, but recognition reliability deteriorates
Solution Approach 1:
The patent uses partial action by selecting and matching only the necessary visible feature points rather than processing all feature points. This reduces the number of comparisons and calculations required, improving recognition speed while maintaining reliability by focusing on the most discriminative visible features.
Solution Approach 2:
The system performs preliminary filtering by checking boarding conditions before conducting face recognition. This preliminary step ensures that recognition is only attempted when conditions are favorable, and the feature point reduction is applied selectively, thereby maintaining reliability while improving speed for appropriate cases.
Data Source
AI summary
An apparatus for recognizing a face is provided. The apparatus includes a camera configured to obtain a partial-face image of a passenger, and a controller configured to learn a boarding condition of a user and recognize the user based on the partial-face image of the passenger when a current boarding condition of the passenger satisfies the learned boarding condition.