Face Recognition Threshold Adjustment for Tracking Loss
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing image recognition systems face challenges in reliably identifying individuals due to changes in facial expressions and orientations, leading to increased time in determining the correct person and potential misrecognition of others as registered individuals.
Innovation Solution
An image processing apparatus with a detection circuit, recognition circuit, tracking circuit, and setting circuit that adjusts threshold values based on the presence or absence of recognized subjects, setting a lower threshold when the subject is not tracked and higher when it is, to enhance recognition accuracy and efficiency.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If a high threshold value is set for individual recognition, then misrecognition is reduced, but recognition time increases and the system may fail to identify the correct person when facial expressions or orientations change
Solution Approach 1:
The patent applies dynamics by making the threshold value changeable based on tracking status. The recognition threshold is set to a first value when the subject is being tracked and a second value (lower than the first) when tracking is lost, allowing the system to adapt between high reliability during stable conditions and faster recognition when conditions change
Solution Approach 2:
The patent implements parameter changes by modifying the recognition threshold parameter according to tracking state. The threshold dynamically switches between two discrete values based on whether the subject region is successfully tracked, enabling the system to balance between recognition accuracy and speed depending on current operational conditions
2Productivity
If a low threshold value is set for individual recognition, then recognition speed increases, but misrecognition of other persons as registered persons increases
Solution Approach 1:
The system dynamically adjusts the threshold based on tracking status, using a lower threshold value only when tracking is lost and subject re-detection is needed, while maintaining a higher threshold during stable tracking to prevent misrecognition
Solution Approach 2:
The recognition threshold parameter is changed based on operational context - a second (lower) value is used when tracking is lost to speed up re-detection, while a first (higher) value is used during normal operation to maintain accuracy and prevent misrecognition
3Reliability
If the threshold value is set to accommodate facial expression and orientation changes, then recognition reliability improves, but the time required to determine the correct person increases
Solution Approach 1:
The patent uses dynamics to adjust the threshold based on tracking status. When tracking is successful despite facial changes, the higher threshold maintains reliability. When tracking is lost, the lower threshold enables faster re-detection, thus resolving the time-reliability tradeoff
Data Source
AI summary
A face detection circuit of an image processing apparatus detects a face region from a video signal. An individual recognition circuit performs individual recognition processing based on a comparison of a similarity between feature information extracted from the face region and registered feature information with a predetermined threshold value. A camera microcomputer tracks a face recognized by the individual recognition circuit and stores information about the tracked face. Further, in a case where the recognized face is not tracked and recognition information about a first subject is not stored, the camera microcomputer sets a first threshold value as a predetermined threshold value for the first subject. On the other hand, in a case where the recognition information about the first subject region is stored, the camera microcomputer sets as the predetermined threshold value for the first subject a second threshold value which is lower than the first threshold value.