Recognition-Avoidance Detection for Occluded Person Images

Resolve Bottlenecks,
Find Innovative Solutions
Generate 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

VSEngineering 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

Engineering Contradiction:
Improverecognition accuracyVSAvoidprivacy violation
Core Design Contradiction:
Measurement precisionVSObject-affected harmful factors

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

Inventive Principle:
Principle #10Preliminary action

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

Inventive Principle:
Principle #24Intermediary (Mediator)

2Reliability

If the system uses full body features for recognition, then recognition reliability is improved, but the ability to respect user privacy preferences deteriorates

Engineering Contradiction:
Improverecognition reliabilityVSAvoidprivacy preference adaptation
Core Design Contradiction:
ReliabilityVSAdaptability or versatility

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

Inventive Principle:
Principle #15Dynamics

3Object-affected harmful factors

If the system analyzes pixel values to detect occlusion, then privacy protection is improved, but processing complexity increases

Engineering Contradiction:
Improveprivacy protectionVSAvoidprocessing complexity
Core Design Contradiction:
Object-affected harmful factorsVSDevice complexity

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

Inventive Principle:
Principle #2Taking out (Extraction)

Data Source

PatentEP3776331B1Detecting actions to discourage recognition
Publication Date: 2025.09.17 GOOGLE LLC
  • EP3776331B1 patent drawingFigure 1
  • EP3776331B1 patent drawingFigure 2
  • EP3776331B1 patent drawingFigure 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.