3D Subject Tracking Using Virtual Camera Distance Images
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing subject tracking technologies struggle in various imaging environments, particularly where subjects' positions change significantly due to varying heights or complex postures, making accurate tracking difficult, especially in sports like keirin and movie captures with wire action.
Innovation Solution
An image processing apparatus that generates and utilizes first and second shape information to set tracking points based on distance information, allowing for consistent identification of subjects even in complex environments by setting a common identifier when distance between tracking points is within a predetermined value, and employs a virtual camera to generate distance images for precise tracking.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If conventional subject tracking methods using predetermined height estimation are used, then tracking can be implemented with simple processing, but tracking accuracy deteriorates in environments with significant height variations such as keirin or wire action movies
Solution Approach 1:
The patent transitions from two-dimensional image plane tracking to three-dimensional space tracking by generating 3D shape information from multiple 2D images. This dimensional change enables accurate tracking of subjects with varying heights and postures by representing their spatial positions and shapes in 3D space, resolving the limitation of conventional 2D tracking methods.
Solution Approach 2:
The patent changes the tracking parameter from fixed height-based estimation to dynamic 3D shape-based tracking. By calculating distance information from virtual camera positions to 3D shape surfaces and identifying tracking points based on these distance variations, the system adapts to subjects with different heights, postures, and positions in three-dimensional space.
2Measurement precision
If multiple imaging apparatuses are used to capture subjects from different angles, then three-dimensional shape information can be generated, but the system complexity and cost increase
Solution Approach 1:
The patent introduces a virtual camera as an intermediary computational model to synthesize multiple real camera viewpoints. By calculating distance information from virtual camera positions to 3D shape surfaces, the system can generate accurate three-dimensional tracking data without requiring physical cameras at every viewpoint, reducing hardware complexity while maintaining measurement precision.
Solution Approach 2:
The patent creates virtual copies of camera viewpoints through computational modeling. Instead of deploying multiple physical imaging apparatuses at every desired angle, the system generates virtual images and distance information from computationally synthesized camera positions, achieving multi-viewpoint 3D tracking with a single physical camera system.
3Reliability
If tracking points are set based on fixed height estimation, then processing is simplified, but tracking reliability deteriorates when subjects change posture or position in three-dimensional space
Solution Approach 1:
The patent implements dynamic tracking point selection that adapts to changing subject postures and positions. Instead of using fixed height-based tracking points, the system continuously calculates distance information from virtual camera positions to 3D shape surfaces and identifies tracking points based on current spatial configuration, ensuring reliable tracking throughout motion sequences with varying postures.
Data Source
AI summary
An image processing apparatus is provided and acquires first shape information and second shape information representing three-dimensional shapes of a subject generated based on captured images at different imaging times, sets a first tracking point at a first imaging time and a second tracking point at a second imaging time based on the first shape information and the second shape information, and sets a common identifier as an identifier set for the second tracking point as the identifier for the first tracking point when a distance between a position of the first tracking point and a position of the second tracking point is less than or equal to a predetermined value.