Object Tracking Apparatus Using Feature Alignment and Area Reuse
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current object tracking methods face challenges in efficiently tracking objects, especially when the object moves or is partially occluded, leading to increased computational resource consumption and potential degradation in 3D image quality due to frequent detector usage.
Innovation Solution
The method involves detecting a target object's initial area, aligning it using feature points, and determining a tracking area based on these points, minimizing detector usage by reusing the tracking area when verified to contain the object, and employing multiple recognizers and aligners trained on different complexities and image qualities to enhance precision and reduce resource consumption.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If the detector is used frequently to track the target object, then the tracking reliability is improved, but the computational resource consumption increases
Solution Approach 1:
The system performs preliminary detection to obtain an initial bounding box and feature points, then uses this information to predefine a tracking area for subsequent frames. This preliminary action allows the system to avoid frequent full-image detection while maintaining tracking reliability, as the tracking area is strategically positioned based on prior detection results.
Solution Approach 2:
The detection process is segmented into two stages: initial detection to establish the tracking area, and subsequent verification within the tracking area only. This segmentation divides the computationally intensive detection task into a one-time full search followed by efficient localized verification, reducing overall computational resource consumption.
2Measurement precision
If the tracking area is dynamically adjusted based on feature points, then the tracking precision is improved, but the device complexity increases
Solution Approach 1:
The system applies different processing qualities to different regions: full-image detection is performed only once to establish the initial tracking area, while subsequent tracking operations are confined to the localized tracking area. This local quality approach improves tracking precision within the region of interest while reducing overall computational complexity.
Solution Approach 2:
The system creates a simplified copy of the detection process that operates only within the predefined tracking area. Instead of repeating the full detection algorithm in every frame, a streamlined verification process is copied and applied locally, maintaining precision while reducing device complexity.
3Measurement precision
If multiple recognizers are used to verify the target object, then the detection accuracy is improved, but the processing time increases
Solution Approach 1:
The system uses multiple recognizers to perform verification, but only within the limited tracking area rather than the entire image. This partial action approach maintains the accuracy benefits of multiple recognizers while significantly reducing processing time by restricting their operation to a small subset of the total image data.
Data Source
AI summary
An object tracking method and apparatus based on object verification are provided. The object tracking method includes detecting a first area that corresponds to a target object in a current frame, aligning the target object with respect to the current frame, and tracking the target object in a next frame based on whether a second area corresponds to the target object as a result of the alignment.