Adaptive Face Search Method for Large-Scale Feature Databases
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Face search methods struggle to efficiently handle large-scale face image feature data, leading to performance issues in matching operations.
Innovation Solution
The method categorizes image feature databases into first-type and second-type based on data volume, using direct matching for first-type databases and index-based matching for second-type databases, reducing the number of features to be matched while ensuring accuracy.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Productivity
If a single face search method is used to match image features one by one, then matching accuracy is maintained, but search efficiency deteriorates when dealing with large-scale face image feature data
Solution Approach 1:
The patent segments the face search process into two distinct methods: a first face search method for small-scale data and a second face search method for large-scale data. This segmentation allows the system to select the appropriate method based on data volume, thereby improving search efficiency without compromising accuracy. The segmentation is implemented through obtaining the quantity of image features, comparing it with a threshold, and selectively executing different search algorithms.
Solution Approach 2:
The patent introduces dynamic adaptability by making the search method selection dependent on the actual data volume. The system dynamically determines which search method to use based on the quantity of image features in the target database, transitioning between different search strategies as conditions change. This dynamic approach ensures optimal performance across varying data scales.
2Reliability
If all image features are matched one by one, then comprehensive search coverage is achieved, but the time consumption increases significantly
Solution Approach 1:
The patent applies partial action by using the second face search method which processes image features in a optimized manner for large-scale data, rather than exhaustively matching every single feature one by one. This partial approach maintains sufficient search coverage while significantly reducing time consumption by leveraging efficient data structures and algorithms appropriate for large datasets.
Solution Approach 2:
The patent changes the operational parameters of the search process based on data volume. When the quantity of image features exceeds the threshold, the system switches to the second search method which employs different matching parameters and algorithms optimized for large-scale data, thereby reducing time consumption while maintaining acceptable search coverage.
3Adaptability or versatility
If a unified face search method is implemented, then system simplicity is maintained, but adaptability to different data scales deteriorates
Solution Approach 1:
The patent creates a universal face search system that can handle both small-scale and large-scale data through a multi-functional architecture. The system includes both the first and second face search methods, along with a selection mechanism that automatically chooses the appropriate method based on data volume. This universal design provides adaptability to different data scales while managing complexity through automated method selection.
Solution Approach 2:
The patent introduces an intermediary component that mediates between the two different search methods and the target database. This intermediary obtains the quantity of image features, compares it with the threshold, and determines which search method to execute. This mediator simplifies the overall system by providing a single entry point that automatically adapts to different data scales without requiring complex manual configuration.
Data Source
AI summary
Disclosed are a face search method and device, an apparatus, and a computer-readable storage medium. According to the data volume of feature information in image feature databases currently to be searched, a user may flexibly select a technique of matching search feature information against feature information in a corresponding feature database, and/or select a technique of first performing initial matching for the search feature information against indices corresponding to feature information in a corresponding feature database, and then performing exact matching for the search feature information against feature information corresponding to a successfully matched index obtained from the initial matching.