Recognition-Avoidance Detection for Occluded Person Images
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Modern computer vision systems can recognize individuals in images or videos by extracting features from the entire body, even when the face is not visible or highly occluded, violating personal privacy.
Innovation Solution
A method and system that analyze pixel values and facial expressions to generate a recognition avoidance signal, determining if a person is attempting to obscure their face or body parts to discourage recognition, and adjust processing accordingly.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If the system extracts features from the entire body for recognition, then recognition accuracy is improved, but personal privacy is compromised
Solution Approach 1:
The system performs preliminary detection of face occlusion actions before executing the recognition process. By analyzing pixel values and generating occlusion signals in advance, the system can determine whether a person is attempting to avoid recognition, and only then proceed with or without recognition based on this preliminary assessment
Solution Approach 2:
The patent introduces an intermediary occlusion detection mechanism between the image input and the recognition process. This intermediary layer analyzes pixel values to detect occlusion signals and determines whether recognition should be performed, acting as a mediator that balances recognition accuracy with privacy protection
2Reliability
If the system uses full body features for recognition, then recognition reliability is improved, but the ability to respect user privacy preferences deteriorates
Solution Approach 1:
The system dynamically adjusts its recognition behavior based on real-time detection of occlusion signals. Rather than using a fixed recognition approach, the system modifies its operation in response to detected user actions, making the recognition process adaptive to user privacy preferences while maintaining reliability when no occlusion is detected
3Object-affected harmful factors
If the system analyzes pixel values to detect occlusion, then privacy protection is improved, but processing complexity increases
Solution Approach 1:
The system extracts only the necessary information for occlusion detection from the image data - specifically analyzing pixel values in the face region to generate occlusion signals. Rather than processing the entire image for recognition, it extracts and analyzes only the relevant occlusion indicators, reducing overall processing complexity while maintaining privacy protection
Data Source
Figure 1
Figure 2
Figure 3A~3B
AI summary
A method to determine whether to perform recognition on an image or a video. The method includes detecting a person in an image or a video by determining a person image region that corresponds to boundaries of the person or a face image region that corresponds to a position of a face of the person. The method further includes person or the position of the face of the person, based on the boundaries of the person or the face image region, to generate a recognition avoidance signal. The method further includes determining whether the recognition avoidance signal indicates that an action was taken to discourage recognition of the person. The method further includes responsive to the recognition avoidance signal indicating that the action was taken to discourage recognition of the person, declining to perform recognition of the person.