Partial Area Feature Selection for Face Recognition Accuracy
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing object identification systems face challenges in maintaining recognition accuracy due to variations in illumination, orientation, and facial expressions, which require a large number of registered images to account for various conditions, leading to increased processing load and decreased user-friendliness.
Innovation Solution
An object identification apparatus that stores feature quantities of partial areas in sample images, selects corresponding sample images based on feature similarities, and uses a discriminator to learn and determine object identity, reducing the need for extensive learning data and processing load by utilizing similarity features.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If many registered images are prepared to account for various conditions (illumination, orientation, facial expressions), then recognition accuracy is improved, but device complexity and processing load increase
Solution Approach 1:
The patent divides the face image into multiple local areas (e.g., eyes, nose, mouth regions) and processes each area separately. By segmenting the face into distinct regions, the system can selectively integrate similarity information from different local areas, reducing the need to process entire high-resolution images under all possible conditions. This segmentation approach maintains recognition accuracy across various illumination and expression conditions while reducing overall processing complexity.
Solution Approach 2:
The patent applies different processing strategies to different local areas of the face based on their specific characteristics. Certain regions (like the eyes or mouth) may be more sensitive to expression changes, while other regions (like the nose) are more stable. By adjusting the weight or processing method for each local area according to its quality and stability, the system achieves robust recognition without uniformly processing all areas with high computational cost.
2Reliability
If many registered images are prepared to account for various conditions, then recognition accuracy is improved, but ease of operation deteriorates
Solution Approach 1:
The patent performs preliminary processing of the registered image by dividing it into multiple local areas and pre-calculating similarity features for each region. This preliminary action allows the system to store compact representations of different facial regions rather than storing numerous complete images under various conditions. When authentication occurs, the system can quickly compare local area features without requiring the user to have registered multiple images, thus improving ease of operation while maintaining accuracy.
3Measurement precision
If similarity features from multiple local areas are integrated, then recognition accuracy is improved, but processing load increases
Solution Approach 1:
The patent selectively integrates similarity information from multiple local areas rather than processing all possible features exhaustively. By focusing on key facial regions and integrating their similarity features with appropriate weighting, the system achieves sufficient identification accuracy without the excessive processing load that would result from analyzing every possible feature across the entire face image under all conditions.
Data Source
AI summary
An object identification apparatus selects, for each partial area of an object in a registered image, a corresponding sample image from sample images based on feature quantities of the partial area, for objects in the registered image, sets a similarity of feature quantities for each partial area between objects of an identical individual and between objects of different individuals based on a similarity related to the selected sample images for each partial area, makes a discriminator learn based on the set similarity, acquires a similarity for each partial area between objects in an input image and the registered image, and determines whether the object in the input image is identical to the object in the registered image based on the acquired similarity and a result of discrimination by the discriminator.