Facial Recognition via High Probability Group Database
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Facial recognition systems face challenges in real-time processing due to large databases and connectivity latency, especially when recognizing individuals at the edge of the Internet, leading to computational bottlenecks and low recognition rates for unseen angles.
Innovation Solution
The implementation of a high probability group (HPG) database that stores likely individuals appearing in video frames, using face detection and tracking algorithms to reduce database searches and enhance recognition with a facial network of related persons, allowing for real-time processing with minimal resources.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If a large database of facial images is used to improve recognition accuracy, then recognition rates improve, but computational complexity and processing time increase
Solution Approach 1:
The patent segments the large facial database into multiple smaller databases organized by geographic location, time period, and other relevant criteria. Each edge device maintains only the subset of facial images relevant to its specific context, transforming one large database into many small, specialized databases that are easier to process while maintaining high recognition accuracy for local scenarios.
Solution Approach 2:
The patent implements local quality by allowing different edge devices to maintain different subsets of facial images based on their specific geographic locations, time periods, and operational contexts. Each device optimizes its database composition locally rather than using a universal database, reducing computational complexity while preserving recognition accuracy for local conditions.
2Measurement precision
If a large database of facial images is stored locally to improve recognition accuracy, then recognition rates improve, but device storage requirements and processing overhead increase
Solution Approach 1:
The patent divides the comprehensive facial database into segmented subsets distributed across multiple edge devices based on geographic regions, time periods, and other relevant dimensions. Each device stores only the necessary portion locally, reducing individual device storage requirements while collectively maintaining comprehensive coverage for accurate recognition.
Solution Approach 2:
The patent introduces additional organizational dimensions (geographic location, time period, device identity) to structure the database distribution. By organizing facial images across multiple dimensions rather than storing all images at each device, the system reduces the quantity of data per device while maintaining comprehensive recognition capability through multi-dimensional access.
3Adaptability or versatility
If cloud connectivity is used to access facial databases remotely, then database accessibility improves, but connectivity latency reduces real-time processing capability
Solution Approach 1:
The patent performs preliminary action by pre-loading and caching relevant facial image subsets into local databases at each edge device before they are needed for real-time processing. This advance preparation eliminates the need for cloud connectivity during critical recognition moments, reducing latency while maintaining database accessibility through local storage.
Solution Approach 2:
The patent introduces local edge devices as intermediaries between the cloud-based comprehensive database and the real-time processing requirements. These intermediaries cache and manage local subsets of facial images, mediating between cloud accessibility and local processing speed to eliminate connectivity latency while preserving database accessibility.
4Productivity
If facial recognition is performed in real-time at the edge with limited resources, then processing speed improves, but recognition accuracy for unseen angles deteriorates
Solution Approach 1:
The patent applies preliminary action by pre-processing and augmenting facial images in the local database with various angles, lighting conditions, and transformations before deployment. This advance preparation ensures that when real-time recognition occurs at the edge with limited resources, the system can accurately recognize faces from unseen angles because multiple angle variations were already prepared in advance.
Solution Approach 2:
The patent implements dynamics by creating adaptive local databases that can be dynamically updated and refined based on local conditions and encountered scenarios. The system adapts to unseen angles and conditions by learning from local data and adjusting the local database composition, maintaining both processing speed and recognition accuracy through dynamic adaptation rather than static pre-processing alone.
Data Source
AI summary
Techniques are provided for facial recognition using a high probability group database and a facial network of related persons. One method comprises maintaining a probability-based database of facial images comprising a subset of individuals from a first database of facial images of a plurality of individuals based on a probability of individuals appearing in sequences of image frames at a given time; applying a face detection algorithm to at least one sequence of image frames to identify one or more faces in the sequences of image frames; maintaining a facial network of related persons associated with the probability-based database by obtaining facial images of one or more additional individuals from the first database that satisfy a predefined related person criteria with respect to individuals identified in at least one sequence of image frames; and applying a facial recognition to at least sequence of image frames using at least the probability-based database and the facial network of related persons to identify individuals in the sequence of image frames.