Face Recognition Similarity Threshold Adaptation
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Face recognition systems face challenges in controlling intrusion risk due to unregistered users, as existing methods lack effective mechanisms to determine similarity thresholds for secure identity verification.
Innovation Solution
A method and apparatus that acquire images of unregistered users, calculate the maximum similarity between their face objects and registered images, and determine a similarity threshold based on a preset condition, ensuring a controlled risk of intrusion by adjusting the ratio of images with high similarity scores.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If a fixed similarity threshold is used for face recognition, then the recognition process is simple and fast, but the intrusion risk cannot be controlled and security is compromised
Solution Approach 1:
The patent performs preliminary actions by acquiring images of unregistered users and calculating their maximum similarities against registered images before actual authentication occurs. This preliminary data collection and analysis enables the system to determine appropriate similarity thresholds in advance, resolving the contradiction between simple fixed thresholds and secure adaptive thresholds.
Solution Approach 2:
The patent implements feedback mechanisms by using the calculated maximum similarities from unregistered users to adjust and determine the similarity threshold. This feedback loop allows the system to continuously improve its security by learning from the distribution of similarity scores and adapting the threshold accordingly, rather than relying on a static predetermined value.
2Reliability
If the similarity threshold is set high to reduce intrusion risk, then security improves, but the recognition pass rate decreases
Solution Approach 1:
The patent applies parameter changes by dynamically adjusting the similarity threshold based on the calculated maximum similarities from unregistered users. Instead of using a fixed threshold, the system modifies the threshold parameter according to the observed distribution of similarity scores, allowing optimal balance between security and recognition accuracy.
Solution Approach 2:
The patent introduces dynamics into the threshold determination process by making the similarity threshold adaptive rather than static. The threshold changes based on real-time analysis of unregistered user images and their similarity scores, enabling the system to respond to varying security requirements and maintain optimal performance.
3Reliability
If images of unregistered users are acquired and analyzed to determine threshold, then security and intrusion control improve, but the system complexity and processing time increase
Solution Approach 1:
The patent performs the time-consuming tasks of acquiring unregistered user images and calculating maximum similarities as preliminary actions during system initialization or periodic updates. By completing these analyses in advance, the actual authentication process can use pre-determined thresholds without requiring real-time analysis, thus reducing processing time during critical authentication operations.
Solution Approach 2:
The patent applies partial action by selectively analyzing only the necessary portions of unregistered user images (focusing on face regions and key features) rather than processing entire images thoroughly. This partial analysis approach reduces processing time while still providing sufficient data to determine accurate similarity thresholds for security purposes.
Data Source
AI summary
The present disclosure discloses a method and apparatus for acquiring information. The method comprises: acquiring images of a plurality of unregistered users, the unregistered users being users having no registered images belonging to the unregistered users in a face recognition system; calculating a maximum similarity corresponding to an image of each unregistered user, the maximum similarity being a maximum similarity among similarities between a face object in the image of the unregistered user and face objects in a plurality of registered images; and determining a similarity threshold corresponding to a preset condition based on the calculated maximum similarity corresponding to the image of the each unregistered user, the preset condition comprising a ratio of a number of images of the unregistered users with corresponding maximum similarities greater than the similarity threshold to a number of the images of the unregistered users being smaller than a ratio threshold.