robust object matching for persistent tracking with heterogeneous features

32
ROBUST OBJECT MATCHING FOR PERSISTENT TRACKING WITH HETEROGENEOUS FEATURES Presented by Enrique G. Ortiz

Upload: ronda

Post on 15-Feb-2016

43 views

Category:

Documents


0 download

DESCRIPTION

Presented by Enrique G. Ortiz. Robust Object matching for persistent tracking with heterogeneous features. System Overview. Prewarping Stage. Approximately match resolutions Define vehicles three directions Chamfer matching for initial translation. Within Sequence Object Mask Generation. - PowerPoint PPT Presentation

TRANSCRIPT

Page 1: Robust Object matching for persistent tracking with heterogeneous features

ROBUST OBJECT MATCHING FOR PERSISTENT TRACKING WITH HETEROGENEOUS FEATURES

Presented by Enrique G. Ortiz

Page 2: Robust Object matching for persistent tracking with heterogeneous features

System Overview

Page 3: Robust Object matching for persistent tracking with heterogeneous features

Prewarping Stage

Approximately match resolutions

Define vehicles three directions

Chamfer matching for initial translation

Page 4: Robust Object matching for persistent tracking with heterogeneous features

Within Sequence Object Mask Generation

Robust Blob Feature Extraction

Blob Tracking via Earth Mover’s Distance

Object Mask Generation

Page 5: Robust Object matching for persistent tracking with heterogeneous features

Blob Detection

Page 6: Robust Object matching for persistent tracking with heterogeneous features

Earth Mover’s Distance

Intuition

Transportation Problem

Page 7: Robust Object matching for persistent tracking with heterogeneous features

Region Matching through EMDDistribution 1 Distribution 2

wi,j

wi,j[R, G, B, X, Y, A]di,j

di,j

di,j fi,j=min(wi,j,w’I,j)

Page 8: Robust Object matching for persistent tracking with heterogeneous features

Object Mask Generation

Page 9: Robust Object matching for persistent tracking with heterogeneous features

Segmentation Result

Page 10: Robust Object matching for persistent tracking with heterogeneous features

Cross Sequence Quasi-Rigid Alignment

Page 11: Robust Object matching for persistent tracking with heterogeneous features

Detect and Classify Edges

Canny Edge Detection

Classify Edges

Page 12: Robust Object matching for persistent tracking with heterogeneous features

Descriptor and Matching

Band Image [Y R G B]

Normalized Correlation

γ u,v( ) =f (x,y) − f[ ]x,y

∑ t(x,y) − t[ ]

f (x,y) − f[ ]2t(x,y) − t[ ]

2

Page 13: Robust Object matching for persistent tracking with heterogeneous features

Matching Result

Page 14: Robust Object matching for persistent tracking with heterogeneous features

Reject Outliers

Outliers Edges with similar appearance False matches Edges with cluttered background

Solution: Impose two rigidity constraints

Page 15: Robust Object matching for persistent tracking with heterogeneous features

Reject Outliers - Affine Rigidity Target depth variation smaller than

range Distances should satisfy 1D affine

model

Page 16: Robust Object matching for persistent tracking with heterogeneous features

Reject Outliers – 3D Line Reconstruction Metadata

Camera Rotation

2D Line Correspondences Camera Translation 3D Line Location

Reject line with large 3D location error

Page 17: Robust Object matching for persistent tracking with heterogeneous features

Outlier Rejection Result

Page 18: Robust Object matching for persistent tracking with heterogeneous features

Extend Line Segments Extend until intersection Modify for equal orientation

Page 19: Robust Object matching for persistent tracking with heterogeneous features

Interpolating Flow Fields

Similarity Transformation

Weight flow less as move away from line

Page 20: Robust Object matching for persistent tracking with heterogeneous features

Incorporation of Points

Harris corner detector Match by correlation Alone insufficient for global motion

model

Page 21: Robust Object matching for persistent tracking with heterogeneous features

“Quasi-Rigid” Alignment Results

Page 22: Robust Object matching for persistent tracking with heterogeneous features

“Quasi-Rigid” Alignment Results

Query Model Warped Query

Page 23: Robust Object matching for persistent tracking with heterogeneous features

Match path to a local distribution of patches within constraints provided by aligned images

Represent patches as oriented energy filter outputs Captures significant features Ignores some illumination effects Each patch captures spatial arrange of edge energy and

orientations within patch Perform nearest-neighbor matching with patches in target

image to account for alignment error Score of the patch is computed as that of the best

matching patch within a small range of translations around the patch

Scores from all local patches aggregated to form overall score between two images

Page 24: Robust Object matching for persistent tracking with heterogeneous features

Matching

Oriented Energy Filter (0, 45, 90, 135)

Page 25: Robust Object matching for persistent tracking with heterogeneous features

Patch Matching

- Spatial Edge Energy- Orientation Weighted sum of correlation score

from all the pixels

Page 26: Robust Object matching for persistent tracking with heterogeneous features

Local Correlation Scores

Global Affine Alignment “Quasi-Rigid” Alignment

Page 27: Robust Object matching for persistent tracking with heterogeneous features

Color and Texture

Local Template Matching Similar to Normalized Correlation

Average Color Similarity Measure the angle between RGB vectors

Page 28: Robust Object matching for persistent tracking with heterogeneous features

Results

Page 29: Robust Object matching for persistent tracking with heterogeneous features

Results

Performance well for sequences up to 30 seconds apart (98.8%)

Does well up to approximately 30° pose change

Page 30: Robust Object matching for persistent tracking with heterogeneous features

2D Affine Alignment + global correlation80% for pose < 33°

Their Method91% for pose < 33°

Page 31: Robust Object matching for persistent tracking with heterogeneous features
Page 32: Robust Object matching for persistent tracking with heterogeneous features

End