Viewpoint Tracking for 3D Display Systems A look at the system proposed by Yusuf Bediz, Gözde...

15
Viewpoint Tracking for 3D Display Systems A look at the system proposed by Yusuf Bediz, Gözde Bozdağı Akar
  • date post

    22-Dec-2015
  • Category

    Documents

  • view

    219
  • download

    2

Transcript of Viewpoint Tracking for 3D Display Systems A look at the system proposed by Yusuf Bediz, Gözde...

Viewpoint Tracking for 3D Display Systems

A look at the system proposed by

Yusuf Bediz, Gözde Bozdağı Akar

Outline

Why viewpoint tracking? Proposed System

Object Detection Tracking Viewpoint Calculation

Results Improvements Conclusion

Why viewpoint tracking?

Parallax Increased accuracy in views Lower cost

Proposed System

Train specialized feature trackers to detect eyes

Determine viewpoint Update view

Object Detection Overview

Training based on positive images of several peoples faces,  negative images of backgrounds

Uses suite of linear classifier such as svms to classify ‘features’ such as eyes or face

Makes use of Adaboost to perform several re-weighted simple linear classifiers

Object Detection Training

Reducing Dimensionality

Two Rectangle MethodLossy CompressionRGB to GrayscaleResolution of the Image Reduction

Object Detection - Classification

Windows of an image are selected and resized2 Rectangle calculation is performedResult is passed into a SVM for ClassificationCascading classification occurs on all windows.

Object Detection - Results

96.8 % Accurate with 6 false detections on a BIOId dataset.

Tracking

Modified Lucas Kanade pyramidal tracking algorithm

Viewpoint Calculation

As only the viewing angle is needed no depth estimation is required.

Results

95+% Accuracy on detection 80-90 ms for detection phase 7-8 ms for tracking phase Overall system runs at 20 fps

Improvements

Allow Y as well as X variation in viewpoint Improved tracking algorithm Improve framerate

Questions?

References

Lucas B D and Kanade T 1981, An iterative image registration technique with an application to stereo vision. Proceedings of Imaging understanding workshop, pp 121--130

J.-Y. Bouguet. “Pyramidal implementation of the Lucas Kanade feature tracker”, OpenCV Documentation, Microprocessor Labs, Intel Corp., 2000

Toyama, K. and Hager, G. D. 1999. Incremental Focus of Attention for Robust Vision-Based Tracking. Int. J. Comput. Vision 35, 1 (Nov. 1999), 45-63. DOI= http://dx.doi.org/10.1023/A:1008159011682

Paul Viola , Michael Jones, “Rapid Object Detection using a Boosted Cascade of Simple Features” 2001 IEEE Computer Society Conference on Computer Vision and Pattern Recognition (CVPR'01) - Volume 1

Qiong Wang, Jingyu Yang, and Wankou Yang, “Face Detection using Rectangle Features and SVM”, International Journal of Intelligent Technology Volume 1 Number 3, 2006