Object Tracking Using Depth and Temporal Proximity
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing image tracking methods, particularly template matching, face challenges in maintaining accurate object tracking when objects are small, partially occluded, or at varying distances, leading to instability and incorrect tracking of similar-colored objects.
Innovation Solution
An image capturing apparatus and method that includes a detection unit to identify candidate objects and a determination unit to assess whether detected objects match the main object based on spatial coordinates, distance, and priority criteria, ensuring accurate tracking by switching between face and body detection when necessary.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Adaptability or versatility
If template matching method is used to track objects, then tracking can be performed when objects are small or not facing front, but tracking accuracy deteriorates due to shape and color changes or similar objects
Solution Approach 1:
The patent segments the tracking process into multiple detection methods (face detection, body detection, thing detection) that operate independently and can be selectively combined. Each detection method targets specific object characteristics, allowing the system to choose the most appropriate method based on current conditions while maintaining overall tracking accuracy.
Solution Approach 2:
The patent dynamically changes detection parameters by switching between different detection algorithms based on object characteristics and tracking conditions. When face detection becomes feasible, the system transitions from body/thing detection to face detection, improving accuracy by adapting the detection parameters to current object visibility and orientation.
2Reliability
If correlation determination is performed based on position and size of estimated regions, then object identity can be determined, but false positives occur when similar objects pass by or faces are detected in background
Solution Approach 1:
The patent introduces depth information as an additional dimension to the correlation determination process. By considering not only two-dimensional position and size but also three-dimensional distance information, the system can distinguish between objects at different depths, eliminating false positives from background faces or similar objects that occur in different spatial planes.
3Reliability
If multiple detection methods are used to improve tracking accuracy, then detection reliability improves, but system complexity increases
Solution Approach 1:
The patent implements a dynamic detection system that adaptively selects and switches between different detection methods based on real-time object characteristics and tracking conditions. Rather than running all detection methods simultaneously, the system dynamically activates only the necessary detection type, reducing computational complexity while maintaining high detection reliability through conditional multi-method usage.
Data Source
AI summary
An apparatus comprises a detection unit detecting an object from an image acquired by a capturing unit; and a determination unit that determines whether or not the object is the same as a main object, wherein the determination unit includes a first determination unit that determines whether or not coordinates of the object candidate and coordinates of the main object satisfy a distance condition; a second determination unit that determines whether or not a difference between the object candidate and the main object is within a predetermined range; and a third determination unit that determines whether or not the first and second determination units have determined in the affirmative a predetermined number of times in a row, and the determination unit determines whether transfer of the main object has occurred based on determination results of the first to third determination units.