This document summarizes different methods for monocular model-based 3D tracking of rigid objects using a survey format. It discusses edge-based methods like RAPiD that match model outlines to straight line segments extracted from images. It also reviews optical flow-based methods that track points between frames and techniques that combine optical flow with edge tracking. Additionally, it examines template matching approaches, interest point detection methods like Harris-Stephen and Shi-Tomasi, interest point matching using correlation windows, and pose estimation by tracking planes defined by sets of points. The document provides an overview of key algorithms for model-based 3D tracking and segmentation.

1. Monocular Model-Based 3D Tracking of Rigid Objects: A Survey2008. 12. 15.백운혁Chapter 4. Natural Features, Model-Based Tracking

3. AgendaMonocular Model-Based 3D Tracking of Rigid Objects : A SurveyChapter 4. Natural Features, Model-Based Tracking4.1. Edge-Based Methods4.2. Optical Flow-Based Methods4.3. Template Matching4.4. Interest Point-Based Methods4.5. Tracking Without 3D Models

4. 4.1 Edge-Based Methodsstraight line segments and to fit the model outlines

5. 4.1.1 RAPiD

6. 4.1.1 RAPiDOriginControl pointControl point in camera coordinatesMotion

7. 4.1.1 RAPiD

8. 4.1.1 RAPiDdistanceis vector made of the distances

9. 4.1.2 Making RAPiD RobustMinimize the distanceControl points lying on the same object edge are grouped into primitives. And a whole primitive can be rejected from the pose estimation.RANSAC methodologyThe number of edge strength maxima visible

10. 4.1.3 Explicit Edge ExtractionThe middle point, the orientation and the length of the segmentOf a model segmentOf a an extracted segmentMahalanobis distanceIs the covariance matrixThe pose is then estimated by minimizing

11. 4.2 Optical Flow-Based MethodsIts corresponding location in the next imageThe projection of a point in an image at time

12. 4.2.1 Using Optical Flow AloneNormal optical flowFor large motionsCauses error accumulation

13. 4.2.2 Combining Optical Flow and EdgesTo avoid error accumulationDepends of the pose and the image spatial gradients at timeIs a vector made of the temporal gradient at the chosen locations

14. 4.3 Template MatchingTo register a 2D template to an image under a family of deformations

15. 4.3.1 2D TrackingTo find the parameters of some deformation That warps a template into the input image is the pseudo-inverse of the Jacobian matrix of computed at

16. 4.4 Interest Point-Based MethodsUse localized featuresRely on matching individual features across images and are therefore easy to robustify against partial occlusions or matching errors

17. 4.4.1 Interest Point DetectionHarris-Stephen detector / Shi-Tomasi detectorThe pixels can be classified from the behavior of the eigen values of The coefficients of are the sums over a windowof the first derivatives and of image intensitieswith respect to pixel coordinates

18. 4.4.2 Interest Point Matchingto use7x7 correlation windowsreject matches for which measure is less than 0.8search of correspondents for a maximum movement of 50 pixelsKanade-Lucas-Tomasi trackerKeep the points that choose each other

19. 4.4.3 Pose Estimation by Tracking PlanesPose Estimation for Planar Structures

20. Thanks for your attention