Person Association in Multi-Camera Image Processing
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing image processing systems for monitoring multiple cameras face errors in associating individuals due to similarities in appearance, making it difficult to determine if persons are the same or different, especially in systems requiring high reliability and consistency.
Innovation Solution
An image processing system that includes input means for receiving camera feeds, registration means for registering persons, and display control means allowing users to confirm if persons in the images are the same or different, using user input to estimate correspondence relationships and improve accuracy through similarity calculations and movement history analysis.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Productivity
If automatic association of persons is performed based on similarity of appearance, then productivity is improved, but reliability deteriorates due to errors in associating persons with similar appearances
Solution Approach 1:
The patent introduces an intermediary verification mechanism where user input serves as a mediator between automatic similarity-based association and final association results. The system presents automatically associated results to users for confirmation, allowing users to correct errors while maintaining the efficiency of automatic processing for clear cases.
Solution Approach 2:
The system implements feedback loops where user corrections to automatic associations are fed back into the system to improve future automatic association accuracy. The system learns from user feedback to refine its similarity calculation algorithms and reduce future errors in associating persons with similar appearances.
2Reliability
If human verification is introduced to improve association accuracy, then reliability is improved, but device complexity and operation difficulty increase
Solution Approach 1:
The system applies partial human verification rather than requiring verification for all cases. It uses similarity threshold calculations to automatically associate persons with high confidence ratios without user intervention, while only presenting cases with lower confidence ratios or ambiguous similarities for user verification, thereby reducing overall system complexity.
Solution Approach 2:
The system dynamically adjusts verification parameters based on similarity calculation results. When similarity ratios exceed certain thresholds, automatic association is enabled without user input. When ratios fall below thresholds or show ambiguous patterns, the system transitions to requiring user verification, effectively changing the operational parameters based on input conditions.
3Reliability
If manual association by users is performed, then reliability is improved, but productivity decreases due to time-consuming verification processes
Solution Approach 1:
The system performs preliminary automatic association based on similarity calculations before presenting results to users. This preliminary action pre-processes the data and identifies likely associations, so users only need to verify or correct results rather than perform complete association analysis from scratch, significantly reducing the time required for manual verification.
Solution Approach 2:
The association process is segmented into automatic processing stages and manual verification stages. The system automatically handles clear-cut cases with high similarity ratios, segmenting them from cases requiring user verification. This segmentation allows the system to maintain high productivity for straightforward cases while ensuring reliability for ambiguous cases.
Data Source
AI summary
Provided are an image processing system, an image processing method, and a program capable of suitably performing the association of a person appearing in a picture in accordance with a highly reliable user input. This image processing system includes: an input unit for receiving an input of pictures captured by multiple video cameras; a person-to-be-tracked registration unit capable of registering one or more persons appearing in the pictures input from the input unit; a moving image display unit for displaying, on a display device, the pictures input by the input unit, and a UI generation unit capable of registering that the person appearing in the displayed pictures and the person registered by the person-to-be-tracked registration unit are the same person, or not the same person.