3dv june 29 2013 eric turner avideh zakhor
Download
1 / 48

Watertight Planar Surface Meshing of Indoor Point-Clouds with Voxel Carving - PowerPoint PPT Presentation


  • 109 Views
  • Uploaded on

EE CS. 3DV - June 29, 2013 Eric Turner Avideh Zakhor. Watertight Planar Surface Meshing of Indoor Point-Clouds with Voxel Carving. University of California, Berkeley. EE CS. Surface Meshing of Buildings. University of California, Berkeley. EE CS. Surface Meshing of Buildings.

loader
I am the owner, or an agent authorized to act on behalf of the owner, of the copyrighted work described.
capcha
Download Presentation

PowerPoint Slideshow about ' Watertight Planar Surface Meshing of Indoor Point-Clouds with Voxel Carving' - jada


An Image/Link below is provided (as is) to download presentation

Download Policy: Content on the Website is provided to you AS IS for your information and personal use and may not be sold / licensed / shared on other websites without getting consent from its author.While downloading, if for some reason you are not able to download a presentation, the publisher may have deleted the file from their server.


- - - - - - - - - - - - - - - - - - - - - - - - - - E N D - - - - - - - - - - - - - - - - - - - - - - - - - -
Presentation Transcript
3dv june 29 2013 eric turner avideh zakhor

EECS

3DV - June 29, 2013

Eric Turner

Avideh Zakhor

Watertight Planar Surface Meshing of Indoor Point-Clouds with Voxel Carving

University of California, Berkeley


Surface meshing of buildings

EECS

Surface Meshing of Buildings

University of California, Berkeley


Surface meshing of buildings1

EECS

Surface Meshing of Buildings

University of California, Berkeley

  • Indoor Navigation

  • Energy Simulation

  • Texture Mapping

    • Virtual Reality

    • Augmented Reality

    • Video Games


Indoor modeling

EECS

Indoor Modeling

University of California, Berkeley

  • Acquisition System


Indoor modeling1

EECS

Indoor Modeling

University of California, Berkeley

  • Point-cloud Generation


Example point cloud

EECS

Example Point-cloud

University of California, Berkeley


Example point cloud1

EECS

Example Point-cloud

University of California, Berkeley

Example Mesh


Motivation

EECS

Motivation

University of California, Berkeley

  • Why is meshing useful?

  • Why do we want it to be planar?

  • Why do we want it to be watertight?


Motivation1

EECS

Motivation

University of California, Berkeley

  • Why is meshing useful?

Example

Point-cloud

  • 45 Million Pts

  • 3.5 GB on disk


Motivation2

EECS

Motivation

University of California, Berkeley

  • Why is meshing useful?

Example Mesh

  • 985,000 Tris

  • 20 MB on disk


Approach

EECS

Approach

University of California, Berkeley

Point

Cloud

Path

Voxels labeled inside / outside

Voxel Carving

Region Growing

Planar geometry fit to voxel boundary

Triangulate Regions

Watertight Mesh

Adaptively sized triangles


Voxel carving

EECS

Voxel Carving

University of California, Berkeley

LiDAR scans

Path of Scanner


Voxel carving1

EECS

Voxel Carving

University of California, Berkeley

  • Trace path of laser through space

LiDAR scans

Path of Scanner


Voxel carving2

EECS

Voxel Carving

University of California, Berkeley

  • Interpolate neighboring scans to define volume

Ray-tracing


Voxel carving3

EECS

Voxel Carving

University of California, Berkeley

  • Define volume with voxels

Carved Voxels


Voxel data structure

EECS

Voxel Data Structure

University of California, Berkeley

  • Only boundary voxels explicitly stored

  • Carving preserves watertightness of volume


EECS

University of California, Berkeley


Plane fitting

EECS

Plane Fitting

University of California, Berkeley

  • Goal: remove discretization artifacts on surface


Plane fitting1

EECS

Plane Fitting

University of California, Berkeley

  • Goal: remove discretization artifacts on surface

  • Combine voxel faces into planar regions


Plane fitting2

EECS

Plane Fitting

University of California, Berkeley

  • Goal: remove discretization artifacts on surface

  • Combine voxel faces into planar regions

Initialize via flood-fill


Plane fitting3

EECS

Plane Fitting

University of California, Berkeley

  • Merge regions by computing best-fit plane


Plane fitting4

EECS

Plane Fitting

University of California, Berkeley

  • Final merged regions


Plane fitting5

EECS

Plane Fitting

University of California, Berkeley

  • Final merged regions

  • Identified locations of planar regions

  • Now need to mesh


Triangulation of regions

EECS

Triangulation of Regions

University of California, Berkeley

  • Example planar region


Triangulation of regions1

EECS

Triangulation of Regions

University of California, Berkeley

  • Example planar region

  • Dominant axis-aligned plane


Triangulation of regions2

EECS

Triangulation of Regions

University of California, Berkeley

  • Triangulate 2D projection


Triangulation of regions3

EECS

Triangulation of Regions

University of California, Berkeley

  • Example


Triangulation of regions4

EECS

Triangulation of Regions

University of California, Berkeley

  • Boundaries snapped to plane intersections


Results

EECS

Results

University of California, Berkeley


Results1

EECS

Results

University of California, Berkeley


EECS

University of California, Berkeley

Constructed mesh of hallway

Texture-mapped mesh of hallway


EECS

University of California, Berkeley


EECS

University of California, Berkeley


EECS

University of California, Berkeley


EECS

University of California, Berkeley


EECS

University of California, Berkeley


EECS

University of California, Berkeley

Mesh, colored by region

Close up of hotel hallway

Viewing triangulation and planar regions

Point-cloud


Large retail shopping center

EECS

Large Retail Shopping Center

University of California, Berkeley

2.7 million triangles from 220 million points

112 m x 78 m


Thank you

EECS

Thank You

University of California, Berkeley


Supplemental

EECS

Supplemental

University of California, Berkeley


Model statistics

EECS

Model Statistics

University of California, Berkeley


Preserving fine detail in voxels

EECS

Preserving Fine Detail in Voxels

University of California, Berkeley

Scan points

Path of scanner


Preserving fine detail in voxels1

EECS

Preserving Fine Detail in Voxels

University of California, Berkeley

Scan points

Lasers

Path of scanner


Preserving fine detail in voxels2

EECS

Preserving Fine Detail in Voxels

University of California, Berkeley

Exterior Space

Interior Space


Preserving fine detail in voxels3

EECS

Preserving Fine Detail in Voxels

University of California, Berkeley

Scan points

Path of scanner

Lasers stopped at any occupied voxel


Preserving fine detail in voxels4

EECS

Preserving Fine Detail in Voxels

University of California, Berkeley

Exterior Space

Interior Space


Preserving fine detail in voxels5

EECS

Preserving Fine Detail in Voxels

University of California, Berkeley

Preserving fine detail


Preserving fine detail in voxels6

EECS

Preserving Fine Detail in Voxels

University of California, Berkeley

Without preserving fine detail


ad