PCA Face Recognition Incremental Eigenspace Update
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Conventional face recognition systems using PCA analysis require retraining and loss of stored data when new images are added, making them inefficient and time-consuming, especially for large collections.
Innovation Solution
A method for combining and updating PCA representations from multiple collections without using original facial image samples, allowing for efficient re-projection of data into a new eigenspace and comparison of facial images within a combined collection.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Quantity of substance
If new images are added to the PCA collection, then the image collection is updated and expanded, but the stored PCA-based data becomes invalid and retraining is required
Solution Approach 1:
The system dynamically adapts the PCA basis vectors by computing incremental updates rather than performing complete retraining. When new images are added, the algorithm computes the difference between the old and new mean faces, then updates the basis vectors incrementally to reflect the changed data distribution, allowing the system to remain accurate without full retraining.
Solution Approach 2:
The system performs preliminary computations by pre-calculating the mean face and establishing initial basis vectors before new data arrives. This preliminary setup allows for efficient incremental updates when new images are added, as the foundation is already in place to accommodate changes without starting from scratch.
2Reliability
If complete retraining is performed when adding new images, then the PCA basis vectors are updated, but stored eigenface data is lost and the process becomes inefficient
Solution Approach 1:
Instead of static complete retraining, the system uses dynamic incremental updates to the basis vectors. The algorithm computes only the necessary changes based on new data, maintaining accuracy while preserving existing eigenface data and avoiding the inefficiency of full retraining cycles.
Solution Approach 2:
The system discards only the specific changes needed (difference in mean faces) rather than discarding all stored data. By recovering and updating only the affected portions of the PCA model, the system maintains reliability while improving productivity through selective rather than complete retraining.
3Quantity of substance
If the number of new images exceeds 20% of the original collection, then the collection is significantly expanded, but the self-consistency of PCA results breaks down
Solution Approach 1:
The system maintains stability through dynamic adaptation of the basis vectors. Even when new images exceed 20% of the original collection, the incremental update algorithm continuously adjusts the mean face and basis vectors to reflect the current data distribution, preserving self-consistency regardless of collection size changes.
4Measurement precision
If mean face variation is significant between collections, then a new set of basis vectors should be obtained, but complete retraining is time consuming
Solution Approach 1:
The system performs partial action by computing only the necessary subset of updates based on mean face variation. Rather than always performing complete retraining, the algorithm selectively updates only when and to the extent that mean face variation indicates it is necessary, reducing time loss while maintaining measurement precision.
Data Source
AI summary
A representation framework is determined in a face recognition method for a first collection of facial images including at least principle component analysis (PCA) features. A representation of said first collection is stored using the representation framework. A modified representation framework is determined based on statistical properties of original facial image samples of a second collection of facial images and the stored representation of the first collection. The first and second collections are combined without using original facial image samples. A representation of the combined image collection (super-collection) is stored using the modified representation framework.