Keypoint Time-Series Reidentification Without Facial Data
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Person re-identification in video analytics systems remains challenging, particularly in multi-camera surveillance, and facial recognition, while effective, raises significant privacy concerns.
Innovation Solution
Represent objects in video data as sets of key points forming timeseries, which serve as unique 'fingerprints' for identification across different video feeds, eliminating the need for facial recognition.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If facial recognition is used for person re-identification, then identification accuracy is improved, but privacy concerns increase
Solution Approach 1:
The patent extracts only the necessary geometric and motion features from video data without capturing or storing facial images. By taking out only the essential identification information (keypoint coordinates and motion patterns) while leaving out the sensitive facial data, the system achieves accurate person re-identification while preserving privacy.
Solution Approach 2:
The patent introduces an intermediary representation layer between video input and identification output. Instead of directly using facial images for identification, the system uses keypoints and timeseries as an intermediary that captures identification-relevant information while discarding privacy-sensitive visual details.
2Object-affected harmful factors
If traditional person re-identification methods are used, then privacy is preserved, but re-identification accuracy decreases
Solution Approach 1:
The patent transitions from spatial-only features to spatio-temporal features by introducing the time dimension through timeseries analysis. This dimensional expansion enables accurate re-identification using privacy-preserving methods by capturing motion patterns and temporal dynamics that are invisible to traditional spatial-only approaches.
Solution Approach 2:
The patent changes the feature representation parameters from visual appearance (facial images) to geometric and temporal parameters (keypoint coordinates and motion patterns). This parameter transformation maintains identification accuracy while fundamentally altering the data type to be privacy-preserving.
3Adaptability or versatility
If facial recognition systems are deployed, then person identification capability is improved, but system complexity increases
Solution Approach 1:
The patent segments the person representation into discrete keypoints rather than processing entire facial images. This segmentation simplifies the data structure and processing requirements while maintaining identification capability, reducing system complexity compared to full facial recognition systems.
Data Source
AI summary
In one embodiment, a device represents each of a plurality of objects depicted in video data captured by a plurality of cameras over time as a set of key points associated with that object. The device forms, for each of the plurality of objects, a set of timeseries of the set of key points associated with that object. The device performs reidentification of a particular one of the plurality of objects across video data captured by two or more of the plurality of cameras by matching sets of timeseries of key points associated with that object derived from video data captured by two or more of the plurality of cameras. The device provides an indication of the reidentification for display to a user.