Face Identification Device Dynamic Feature Registration
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing face identification techniques suffer from reduced accuracy due to changes in facial expressions, facing directions, and lighting conditions, and require significant processing load for database updates.
Innovation Solution
An identification device and method that extracts feature data from multiple frames of image data, tracks identical face areas, and registers additional feature data when identification fails in subsequent frames, improving accuracy and reducing processing load.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If feature data is registered under various conditions (facial expressions, lighting, facing directions) to improve identification accuracy, then identification accuracy improves, but processing load increases
Solution Approach 1:
The system automatically determines whether additional registration is needed by comparing identification results across frames, and performs registration only when necessary. The tracking unit and registration unit work autonomously to monitor identification stability and update the database selectively, eliminating the need for manual intervention or continuous full database updates.
Solution Approach 2:
The system changes the registration parameter from continuous updating to conditional updating based on identification stability. By monitoring whether identification results remain consistent across multiple frames under varying conditions (lighting, expression, angle), the system dynamically adjusts when to register additional feature data, optimizing both accuracy and processing efficiency.
2Measurement precision
If database update processing is performed whenever a face image is picked up to improve identification accuracy, then identification accuracy improves, but processing load increases
Solution Approach 1:
The system autonomously monitors identification results across multiple frames and automatically determines when database updates are necessary. The tracking unit continuously tracks face areas and the registration unit selectively updates the database only when identification instability is detected, eliminating the need for manual update triggers or continuous processing.
Solution Approach 2:
Instead of continuous or frequent database updates, the system performs periodic checks by comparing identification results across multiple frames. Updates are triggered only when a change in identification result is detected, creating a rhythm of processing that balances accuracy improvement with computational efficiency.
3Measurement precision
If feature data is extracted and registered for each frame to improve identification accuracy, then identification accuracy improves, but processing load increases
Solution Approach 1:
The system automatically monitors identification results across frames and selectively extracts feature data only when necessary. The tracking unit identifies stable face areas and the registration unit extracts and registers feature data only when identification instability is detected, eliminating redundant processing of every frame.
Solution Approach 2:
Instead of processing every frame equally, the system applies partial processing by extracting and registering feature data only for frames where identification instability is detected. This selective approach processes only the necessary portion of frames, reducing overall processing load while maintaining accuracy.
Data Source
AI summary
An identification device capable of improving identification accuracy. The identification device performs identification according to a face area contained in image data. Feature data is extracted from a face area in each of frames of image data. The extracted feature data is registered in a person database section. Identification is performed through comparison between the feature data registered in the person database section and the extracted feature data. A tracking section identifies an identical face area in consecutive frames. If a face area is identified in a first frame, but a face area in a second frame following the first frame, which is identified by the tracking section as identical to the identified face area in the first frame, is not identified, the extracted feature data associated with the face area in the second frame is registered as additional feature data in the person database section.