thesis by: ran zask 105221 asian institute of technology umm not sure.. where are you?

18
Thesis by: Ran Zask 105221 Asian Institute of Technology Umm not sure.. Where are you?

Upload: paula-wright

Post on 04-Jan-2016

215 views

Category:

Documents


0 download

TRANSCRIPT

Page 1: Thesis by: Ran Zask 105221 Asian Institute of Technology Umm not sure.. Where are you?

Thesis by: Ran Zask 105221

Asian Institute of Technology

Umm not

sure..

Where are

you?

Page 2: Thesis by: Ran Zask 105221 Asian Institute of Technology Umm not sure.. Where are you?

Agenda Background

Problem statement

Related work

Our algorithm

o 3D modeling

o Texture mapping

Results

Problems & limitations

Conclusions & Recommendations

Page 3: Thesis by: Ran Zask 105221 Asian Institute of Technology Umm not sure.. Where are you?

BackgroundSome robot applications require human operator (e.g.

search & rescue)

Human operator needs to maintain situation

awareness in order to control the robot.

Current means of maintaining situation awareness are

poor (2D maps, live video)

3D modeling and visualization is needed.

Page 4: Thesis by: Ran Zask 105221 Asian Institute of Technology Umm not sure.. Where are you?

Problem statementReal-time SLAM implementations do not

estimate surfaces of the environment and

using expensive sensors (sonar, laser)

Structure From Motion is too heavy and done

offline.

Real-time 3D modeling and visualization is

feasible and needed.

Page 5: Thesis by: Ran Zask 105221 Asian Institute of Technology Umm not sure.. Where are you?

Related workPollefeys et al. (2004) : point-based modeling

o Monocular camera

o 3D reconstruction from feature points

o Bundle adjustment

o Rectification, Dense stereo

Offline algorithm, Massive computation

Similar application: Microsoft Photosynth

Page 6: Thesis by: Ran Zask 105221 Asian Institute of Technology Umm not sure.. Where are you?

Related work (cont.)Johnson and Kang (1999) : grid-based

modeling

o Omni-directional stereo

o 2D Delauney triangulation

o Iterative closest point (ICP)

oMesh registration

Massive computation

Page 7: Thesis by: Ran Zask 105221 Asian Institute of Technology Umm not sure.. Where are you?

The algorithmSingle camera, calibratedRequires SLAM (point-based)Uses 3D points and occupancy gridsOnline algorithmCreates 3D models with texture mappingProvides low metric accuracyLess noise between grid cells

Page 8: Thesis by: Ran Zask 105221 Asian Institute of Technology Umm not sure.. Where are you?

The algorithmUses 2 occupancy grids:Local grid

o Recreated each frame

o Contains ‘fullness’ seen at the current frame

o Incrementally added to the global grid.Global grid

o Contains the incremental ‘fullness’

o Used for Iso-surface calculation

Page 9: Thesis by: Ran Zask 105221 Asian Institute of Technology Umm not sure.. Where are you?

The algorithm (3D modeling)For each image: Input: new 3D point set (from SLAM)Reset local gridAssociate 3D points with cells in the local gridIgnore cells associated by few pointsAdd camera-center as an occupied cellFill-in the grid (convex-hull)Merge local grid into global gridIso-surface the global grid -> last model

Page 10: Thesis by: Ran Zask 105221 Asian Institute of Technology Umm not sure.. Where are you?
Page 11: Thesis by: Ran Zask 105221 Asian Institute of Technology Umm not sure.. Where are you?

The algorithm (texture mapping)For each image : (After having the latest model)Classify triangles as: “new,” “old,” and

“expired”Project “new” triangles onto current frame ->

textmapsRemove “expired” triangles

Page 12: Thesis by: Ran Zask 105221 Asian Institute of Technology Umm not sure.. Where are you?

ResultsModeling accuracy is satisfyingTexture mapping

Tessellation within a patch: perfect

Tessellation between patches: Inaccurate

The more images are closer to each other, the better the modeling and the texture mapping

“patch” – collection of several triangles of the same image

Page 13: Thesis by: Ran Zask 105221 Asian Institute of Technology Umm not sure.. Where are you?

Modeling results

Page 14: Thesis by: Ran Zask 105221 Asian Institute of Technology Umm not sure.. Where are you?

Modeling results

Page 15: Thesis by: Ran Zask 105221 Asian Institute of Technology Umm not sure.. Where are you?

Modeling results

Page 16: Thesis by: Ran Zask 105221 Asian Institute of Technology Umm not sure.. Where are you?

SummaryWe created an online algorithm for modeling

with texture, all from a single camera.

Low computation is required per image

This should allow short baseline -> good

results

Real time implementation of the system is

feasible.

Page 17: Thesis by: Ran Zask 105221 Asian Institute of Technology Umm not sure.. Where are you?

Future workCreate a real-time system

Create 3D tools for the human operator to

easily ‘walk’ in the model and navigate the

robot.

Improving the algorithm even more:

performing iso-surface on tighter grids

Page 18: Thesis by: Ran Zask 105221 Asian Institute of Technology Umm not sure.. Where are you?