Subject Tracking via Image Area Segmentation and Similarity Matching
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing subject tracking techniques, such as those based on Japanese Patent Application Laid-Open Publication No. 2004-171490, face accuracy issues when a subject, like a face, changes angle or orientation, leading to difficulties in tracking across multiple images.
Innovation Solution
A subject tracking apparatus and method that divides image data into multiple areas, specifies corresponding areas in each frame based on characteristic values, and tracks these areas to maintain accurate subject detection even with changes in angle or orientation.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Productivity
If template matching is used for subject detection, then operation quantity is reduced by searching near the last detected position, but tracking accuracy deteriorates when the subject changes angle or orientation
Solution Approach 1:
The patent divides the subject area into multiple first areas and divides the search area into multiple second areas. This segmentation allows the system to track different parts of the subject independently, maintaining accuracy even when the subject changes orientation. Each divided area can be matched separately, preventing complete tracking failure when parts of the subject move out of the original search range.
Solution Approach 2:
The patent expands the search from a single positional parameter to multiple dimensions by dividing areas into multiple segments. Instead of searching in one location, the system searches across multiple divided areas simultaneously, adding spatial dimensionality to the tracking approach. This allows the system to capture subject movement in various directions and orientations.
2Speed
If the search area is limited to the vicinity of the last detected position, then detection speed is improved, but the system fails to track subjects that turn or change orientation significantly
Solution Approach 1:
By segmenting the search area into multiple second areas corresponding to the divided first areas, the system maintains detection speed while improving reliability. Each segmented area can be processed independently and in parallel, preserving speed, while the collective coverage of all segments ensures the subject remains tracked even during orientation changes.
Solution Approach 2:
The system performs preliminary division of both the subject area and search area into multiple segments before actual tracking begins. This preliminary segmentation prepares the system to handle various subject movements, allowing quick response when the subject changes orientation without requiring complete re-detection.
3Device complexity
If single-area template matching is used, then the process is simple, but accuracy is reduced when subject orientation changes
Solution Approach 1:
The patent applies segmentation to both the subject area (dividing into first areas) and the search area (dividing into second areas). This creates a systematic multi-area matching process that improves accuracy through divided attention to different subject parts, while maintaining manageable complexity through structured organization of the division and matching operations.
Solution Approach 2:
The system applies local quality by treating different divided areas with specific matching operations tailored to their local characteristics. Each first area is matched with corresponding second areas based on local template characteristics, allowing the system to adapt to local orientation changes in different parts of the subject rather than requiring uniform handling of the entire subject.
Data Source
AI summary
A subject tracking apparatus, including: a storage section that stores image data of a subject; a division section that divides the image data of the subject into a plurality of first areas; an imaging section that sequentially generates image frames; a specification section that specifies a second area similar to one of the first areas, the second area being included in each of the image frames sequentially generated, based on a characteristic value of one of the first areas into which the image data of the subject is divided by the division section; and a tracking section that tracks an image area as the subject in each of the image frames sequentially generated, wherein the image area is determined based on the second area specified by the specification section.