Video Image Person Re-identification System
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing systems for wide-area monitoring using video images from multiple cameras face challenges in accurately re-identifying individuals without human intervention, leading to potential errors and complex operations.
Innovation Solution
An image processing system that includes input means for accepting video images from multiple cameras, display control means for displaying video images and associating person images, and registering means for simplifying the re-identification process by displaying person images related to registered individuals in the vicinity of the video image.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If person re-identification is performed manually by human initiative, then identification accuracy is improved, but operational complexity increases
Solution Approach 1:
The system performs automatic person re-identification by itself without requiring manual human intervention. The registration unit automatically registers persons and the display control unit automatically displays associated person images, enabling the system to serve itself in the re-identification process while maintaining high accuracy through automated image matching and comparison algorithms
Solution Approach 2:
The system introduces an intermediary automated processing layer between manual operations and final identification results. The registration unit and display control unit act as intermediaries that automatically handle the complex tasks of person registration, image association, and display generation, thereby reducing operational complexity while preserving identification accuracy
2Ease of operation
If automatic person re-identification is performed using similarity-based methods, then operational complexity is reduced, but identification accuracy deteriorates due to errors
Solution Approach 1:
The system incorporates feedback mechanisms where the display control unit presents associated person images to operators for verification. This feedback loop allows operators to review and confirm automatic matching results, thereby maintaining high identification accuracy while keeping operational complexity low through automated preprocessing and candidate generation
Solution Approach 2:
The registration unit performs preliminary actions by pre-registering persons and pre-associating their images before actual re-identification tasks. This preliminary preparation stores reference data and association relationships in advance, enabling rapid and accurate automatic matching during operation without requiring complex real-time processing
3Reliability
If multiple person images are displayed for re-identification, then identification completeness is improved, but display complexity increases
Solution Approach 1:
The display control unit applies local quality by selectively displaying person images based on their relevance and association strength with the target person. Instead of uniformly displaying all registered person images, the system prioritizes and highlights locally relevant images in the vicinity of the video image, reducing display complexity while maintaining identification completeness
Data Source
AI summary
An image processing system, an image processing method, and a program capable of implementing an association of a person appearing in a video image through a simple operation are provided. The image processing system includes an input device which accepts input of video images captured by a plurality of video cameras, a display screen generating unit which causes a display device to display at least one video image among the video images inputted from the input device, and a tracked person registering unit which is capable of registering one or more persons appearing in the video image displayed by the display device. When a person appears in the video image displayed by the display device, the display screen generating unit selectably displays person images of one or more persons, which are associable with the person appearing in the video image and which are registered by the tracked person registering unit, in a vicinity of the video image.