Face Dictionary Feature Filtering for Recognition Accuracy
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing face recognition systems suffer from incorrect and erroneous determinations, leading to impaired accuracy due to exceptional feature registration and inclusion of irrelevant objects in the face dictionary, resulting in multiple identities for the same person and registration of non-target objects.
Innovation Solution
An image processing technique that extracts feature information, registers it in a dictionary, and uses a similarity determination unit to evaluate and filter feature information based on a predetermined evaluation criterion, ensuring only relevant information is used for object similarity determination.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Quantity of substance
If all extracted feature information is registered in the dictionary, then the quantity of registered data increases, but the determination accuracy deteriorates due to incorrect and erroneous registrations
Solution Approach 1:
The patent extracts and removes incorrect and erroneous feature information from the registered data set. The extraction unit identifies feature information that does not satisfy predetermined evaluation criteria, and the removal unit deletes these problematic entries from the dictionary, thereby maintaining high determination accuracy while preserving the quantity of valid registered data
Solution Approach 2:
The patent changes the parameter of data quality by introducing evaluation criteria that assess the correctness and relevance of extracted feature information. By filtering registered data based on these criteria, the system transforms the registered data set from containing both correct and incorrect entries to containing only validated, high-quality feature information
2Quantity of substance
If feature information from exceptional capturing is registered, then the quantity of registered data increases, but multiple identities are generated for the same person
Solution Approach 1:
The patent implements a feedback mechanism where the similarity determination unit continuously evaluates registered feature information against evaluation criteria. When exceptional feature information that could create multiple identities is detected, the system provides feedback to the removal unit, which then deletes the problematic entries, preventing identity fragmentation while maintaining comprehensive data registration
3Quantity of substance
If objects other than target objects are registered in the dictionary, then the quantity of registered data increases, but erroneous determination occurs
Solution Approach 1:
The patent performs preliminary action by evaluating feature information before it is registered in the dictionary. The extraction unit assesses whether extracted feature information satisfies predetermined evaluation criteria related to target object characteristics. Only feature information that passes this preliminary evaluation is registered, preventing erroneous determination while maintaining comprehensive registration of valid objects
Data Source
AI summary
An apparatus extracts feature information from an object of image data. The apparatus registers the extracted feature information in a dictionary. The apparatus refers to the dictionary and determines a similarity between feature information registered in the dictionary and the extracted feature information. The apparatus does not use, of feature information to be registered in the dictionary, feature information not satisfying a predetermined evaluation criterion in similarity determination.