Object Recognition Apparatus Grouping Feature Data for Accuracy
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing object recognition systems face a decrease in authentication accuracy when the number of feature information for the same object increases in dictionary data, due to shifts in subregions and changes in facial features, leading to recognition errors.
Innovation Solution
An object recognition apparatus that groups feature information from multiple images of the same object based on similarity and registers new feature information accordingly, either associating it with existing groups or creating new groups to maintain accurate recognition.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If multiple face images are stored in dictionary data to improve recognition accuracy under different conditions, then recognition accuracy improves, but the processing time prolongs and the burden on users increases
Solution Approach 1:
The face image is divided into multiple subregions (e.g., left eye, right eye, nose, mouth) and feature information is extracted from each subregion separately. This segmentation allows the system to process only relevant portions of the face image, reducing the overall processing time while maintaining recognition accuracy across different conditions.
2Adaptability or versatility
If the number of registered feature information for the same object increases, then the recognition data becomes more comprehensive, but recognition errors increase due to subregion shifts and feature changes
Solution Approach 1:
Different processing strategies are applied to different subregions based on their local characteristics. Each subregion's feature information is processed according to its specific properties, allowing the system to maintain high recognition accuracy even when multiple images are registered, by focusing on stable local features rather than global variations.
Data Source
AI summary
In a personal authentication apparatus that compares input feature information with feature information stored in advance as dictionary data, thereby calculating a similarity and recognition a person, when additionally storing feature information in the dictionary data, the feature information is compared with the feature information of the same person already stored in the dictionary data. Pieces of feature information are put into groups for the same person based on the similarities and stored in the dictionary data.