Multi-Stage Object Tracking via Biometric Association
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Conventional moving object tracking devices face challenges in associating objects across cameras, especially when objects undergo changes in clothing or environmental conditions, and struggle to maintain tracking over long periods due to reliance on appearance-based methods.
Innovation Solution
The system employs a moving object tracking device with a CPU-controlled configuration that includes an acquirer, tracker, first associator, and second associator, using multi-channel image processing, biometric authentication, and metric learning to generate and associate tracking information across cameras, ensuring accurate object tracking despite changes in appearance and environment.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Device complexity
If appearance-based tracking is used to associate moving objects across cameras, then tracking can be performed with simple methods, but tracking accuracy deteriorates when objects undergo changes in clothing or environmental conditions
Solution Approach 1:
The system changes the parameters used for tracking from simple appearance-based features to multiple parameters including time difference, appearance similarity, and biometric data. This allows the system to adapt to changes in clothing and environmental conditions while maintaining tracking accuracy.
Solution Approach 2:
The system combines multiple types of tracking information (appearance-based tracking, time difference analysis, and biometric authentication results) into a composite tracking approach. This multi-component method resolves the contradiction by integrating simple and complex methods to achieve both ease of operation and high accuracy.
2Adaptability or versatility
If face recognition is used to track pedestrians, then association is possible despite appearance variations, but tracking is limited to sections where the face can be seen
Solution Approach 1:
The system makes the tracking method universal by applying multiple approaches (appearance-based tracking, time difference analysis, and biometric authentication) to different situations. This allows the system to function effectively whether the pedestrian's face is visible or not, extending tracking capability to all scenarios.
Solution Approach 2:
The system introduces time difference analysis as an intermediary method that bridges the gap when face recognition cannot be applied. By using appearance similarity combined with time difference constraints, the system can track pedestrians even when their face is not visible, thus extending tracking duration.
3Productivity
If appearance-based tracking is used, then tracking can be performed over short periods, but tracking cannot be maintained over long periods when appearance changes
Solution Approach 1:
The system performs preliminary action by collecting and analyzing multiple parameters (appearance features, time differences, and biometric data) in advance. This preparation enables the system to maintain tracking over long periods by having multiple criteria ready to handle appearance changes before they occur.
Solution Approach 2:
The system implements feedback mechanisms by continuously monitoring tracking results and adjusting the weighting of different parameters based on current conditions. When appearance changes are detected, the system feedback-adjusts to rely more on time difference and biometric information, thereby maintaining tracking duration while preserving tracking speed.
Data Source
Figure 1~2
Figure 3
Figure 4~5
AI summary
According to an arrangement, a moving object tracking device (20) includes an acquirer (211), a tracker (212), a first associator (213), a second associator (214), and an output unit (215). The acquirer (211) is configured to acquire images. The tracker (212) is configured to generate pieces of tracking information indicating information obtained by tracking a moving object included in the images. The first associator (213) is configured to generate first associated tracking information obtained by first association targeting the pieces of tracking information having a mutual time difference equal to or smaller than a threshold. The second associator (214) is configured to generate second associated tracking information obtained by second association targeting the first associated tracking information and the tracking information, not associated by the first association, based on authentication information for identifying the moving object. The output unit (215) is configured to output the second associated tracking information.