Confidence-Guided Frame Skipping to Enhance Object Tracking Speed
Author Information
Author(s): Lee Yun Gu
Primary Institution: Kwangwoon University
Hypothesis
Can a confidence-guided frame skipping method improve the speed of object tracking in computer vision?
Conclusion
The proposed method significantly enhances object tracking speed while maintaining accuracy by using a lightweight tracking algorithm and invoking a robust algorithm only when necessary.
Supporting Evidence
- The proposed method achieved a processing speed of 428.2 FPS with a maximum SN of 10.
- Using the confidence level evaluation improved tracking accuracy and robustness compared to methods without it.
- The integration of the proposed method with existing algorithms demonstrated significant speed improvements while maintaining similar accuracy.
Takeaway
This study found a way to make tracking objects in videos faster by using a simple method most of the time and only using a complicated method when needed.
Methodology
The study used a lightweight tracking method based on a block-matching algorithm and evaluated its performance against a robust algorithm, integrating both to enhance tracking speed.
Limitations
The lightweight tracking method may struggle in challenging scenarios such as severe occlusion or rapid object appearance changes.
Digital Object Identifier (DOI)
Want to read the original?
Access the complete publication on the publisher's website