Face Authentication Watchlist Segmentation for False Detection Control
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
In video surveillance systems, the number of false person detection cases increases as the number of person candidates in the watchlist grows, making it challenging to maintain recognition accuracy while authenticating appropriate individuals.
Innovation Solution
A face authentication system with two watchlists and a watchlist updating unit that compares face feature information from input images against registered data, prioritizing matches and updating priority degrees to manage the lists effectively, ensuring accurate authentication.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Adaptability or versatility
If the number of person candidates in the watchlist is increased, then the coverage of person detection is improved, but the false detection rate increases
Solution Approach 1:
The watchlist is divided into multiple lists (first watchlist, second watchlist, third watchlist) with different priorities and purposes. The first watchlist contains high-priority candidates for immediate authentication, the second watchlist contains medium-priority candidates, and the third watchlist contains low-priority candidates. This segmentation allows the system to maintain comprehensive coverage while managing false detection rates through prioritized processing.
Solution Approach 2:
Different watchlists are assigned different quality characteristics - the first watchlist focuses on high-confidence matches with strict authentication, while subsequent watchlists handle lower-priority candidates with adjusted authentication thresholds. This local differentiation of quality allows optimized performance for different candidate groups.
2Reliability
If the number of person candidates in the watchlist is reduced, then the false detection rate is suppressed, but the number of undetected persons increases
Solution Approach 1:
By segmenting the watchlist into multiple priority levels, the system can process candidates in a staged manner. The first watchlist handles high-priority candidates that require immediate attention, while the second and third watchlists capture additional candidates with adjusted processing parameters, thereby maintaining detection productivity while controlling false positives.
Solution Approach 2:
The system dynamically adjusts authentication parameters and processing priorities based on the watchlist structure. Different authentication thresholds and comparison methods are applied to different watchlists, allowing the system to adapt its behavior to maintain both low false detection rates and high detection productivity.
3Measurement precision
If multiple watchlists are used with different priorities, then the authentication accuracy is maintained, but the system complexity increases
Solution Approach 1:
The system segments the authentication process into multiple stages corresponding to different watchlists. Each watchlist operates with specific authentication parameters optimized for its purpose, allowing high accuracy for different candidate types while managing complexity through structured organization rather than monolithic processing.
Solution Approach 2:
A controller coordinates the multiple watchlists and authentication apparatuses, managing the complexity by providing centralized control logic that orchestrates the multi-list processing. The controller assigns candidates to appropriate watchlists and coordinates authentication operations, abstracting the complexity from individual processing units.
Data Source
AI summary
A first authentication apparatus executes authentication of a person captured in an input image by comparing face feature information of the person being acquired from an input image, with registered face feature information in a first watchlist. A second authentication apparatus executes authentication of a person unauthenticated with the first watchlist by comparing face feature information of the face feature information for which matching registered face feature information does not exist in the first watchlist, with registered face feature information in a second watchlist. A host server includes a watchlist updating unit that updates registered face feature information in the first watchlist and registered face feature information in the second watchlist, based on at least one of a priority degree for each piece of registered face feature information in the first watchlist and a priority degree for each piece of registered face feature information in the second watchlist.