Multi-camera Person Re-identification via Global Tracks
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Retailers face challenges in understanding customer shopping behaviors in-store, as in-store interaction data is often limited and not easily analyzable for business or security purposes, unlike online data.
Innovation Solution
A system for person re-identification and tracking across multiple cameras, which links video data from different perspectives to track individuals' trajectories, allowing for analysis of in-store behavior and optimization of retail operations and security.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Quantity of substance
If multiple cameras are deployed to capture customer behavior data, then the quantity and quality of shopping behavior information is improved, but the device complexity and data processing burden increases
Solution Approach 1:
The system divides the complex multi-camera tracking problem into manageable segments by creating camera-specific tracks for individual cameras and global tracks that aggregate information across cameras. This segmentation allows independent processing of each camera's data while maintaining overall coordination through the global track structure.
Solution Approach 2:
The patent introduces global tracks as an intermediary data structure that mediates between multiple camera-specific tracks. These global tracks serve as a coordination layer that links detections from different cameras, enabling cross-camera association without requiring direct complex interactions between all camera systems.
2Measurement precision
If camera-specific tracks are maintained for each camera, then the measurement precision of individual camera detection is preserved, but the loss of information occurs when linking across multiple cameras
Solution Approach 1:
The system merges camera-specific tracks into global tracks by establishing links between detections from different cameras that correspond to the same person. This merging process combines information from multiple sources while preserving the precision of individual camera detections through the linking mechanism.
Solution Approach 2:
The system uses feedback mechanisms where global tracks provide contextual information back to camera-specific tracks, allowing the system to refine detections and reduce information loss. The expiration time reset mechanism provides feedback that helps maintain accurate tracking state across camera transitions.
3Reliability
If tracking is maintained indefinitely to ensure continuous identification, then the reliability of person re-identification is improved, but the loss of time and computational resources increases
Solution Approach 1:
The system implements periodic action through expiration times for global tracks. Instead of maintaining tracks indefinitely, the system periodically evaluates whether tracks should be maintained by checking if they have been updated within their expiration period. This periodic review mechanism balances reliability with computational efficiency.
Solution Approach 2:
The system dynamically changes the expiration time parameter for global tracks based on activity. When a global track is updated with new camera-specific track information, the expiration time is reset, allowing the track to remain active. This parameter change mechanism ensures tracks are maintained only as long as needed for reliable re-identification.
Data Source
AI summary
A system may track a person at a location across multiple cameras in real time, while maintaining identification of the person through their visit. The person may not be personally identified; however, the person's presence in video frames captured by different cameras may be linked together. The linked frames may be evaluated by one or more analytics applications. For example, an analytics application may determine a trajectory of a person through a location and plot an overlay of that person's trip through the location on the map. This information may be used, for example, to optimize store layout, asset protection purposes, for counting a number of people at the location at a given time (e.g., for understanding busy hours of the location), determining whether social distancing or other health and/or safety measures may be in compliance, etc.