Difference between revisions of "TLV3D"

From TAMI
Jump to navigationJump to search
(Created page with "=TLV3D= The idea - to map TLV in 3D using photogeometry. We have the data on official TAMI superclusters and release it at will. == how to == The 3d scanners such as mant...")
 
 
(22 intermediate revisions by 3 users not shown)
Line 1: Line 1:
=TLV3D=
=TLV3D=
The idea - to map TLV in 3D using photogeometry.   
The idea - to map TLV in 3D using any means possible.   


We have the data on official TAMI superclusters and release it at will.
We maintain the data on official TAMI superclusters and release it as emperors of the digital realm.
 
Alon translation: open data map of Tel Aviv (same license as openstreetmaps).
 
== what already exists ==
=== Under open licenses (CC / other) ===
http://openslam.org/ seems to be parking space for slam projects - so far haven't found any full solutions here.
 
point cloud library http://docs.pointclouds.org/trunk/a02944.html
=== Other not open licenses ===
[http://metroscape3d.com/MetroscapePage.html metroscape] has 3d models of a few tens of  bldgs in tlv
* requires windows/mac for the plugin, and no download link for any model (STL/other). -- Alon
googleearth has also a few tens of tlv buildings
 
[[File:googeearth.jpg|400px]]
 
will google come and nail it all down ? if we scoop them we get to ride in the google helicopter.


== how to ==
== how to ==
Line 19: Line 36:


=== structured light ===
=== structured light ===
try to obtain 3d info from a large-ified kinect hack
try to obtain 3d info from a large-ified kinect hack<br>
 
it is possible to build a massive structured light rig, using a 3000+ ansi projector and a DSLR.<br>
see here > https://code.google.com/p/structured-light/
 
15000ansi projector - its 100nis per 1000 lumans per day
i tried that method and it works well, but you can't be moving to much.
you need to hit the same geometry 3 times with different "binary" patterns.
 
=== stereo (camera + laser, use parallex) maker scanner ===
an opensource diy scanner [http://www.makerscanner.com/ here]
 
=== Time of Flight ===
Cheap LiDARs (relatively) @[http://www.hokuyo-aut.jp/02sensor/07scanner/urg_04lx_ug01.html Hokuyo]
* 10m distance (this model, there are others, maybe 30m? also indoors/outdoors differs)
 
[http://www.st.com/st-web-ui/static/active/en/resource/technical/document/data_brief/DM00078006.pdf?s_searchtype=keyword VL6180] by ST.
 
=== Structure from Motion (SfM) ===
[http://ccwu.me/vsfm/ Wikipedia] article.
 
Not necessarily with an IMU, just a camera.
 
* VisualSFM http://ccwu.me/vsfm/
* Bundler http://ccwu.me/vsfm/
* libmv, looks defunct http://ccwu.me/vsfm/
* Python photogrametry toolbox http://www.arc-team.homelinux.com/arcteam/ppt.php
* open multiple view geometry library http://ccwu.me/vsfm/
 
= Hardware =
== Cameras ==
* http://www.mightexsystems.com/family_info.php?cPath=1_252_273&categories_id=273
* http://www.ptgreystore.com/ - mighty expensive (650$ for USB 3.0 camera, but 64x64 350 fps is interesting - although not for this :)
* sphirical cameras, like [http://www.ptgrey.com/products/ladybug2/ladybug2_360_video_camera.asp ladybug] give 360 video feed
 
= Map Stitching =
* http://mapknitter.org/
 
 
= Monocular Slam =
Monocular is a code word for a single camera, here is a list of algorithms, software has still not been tested.
 
Input is a bunch of camera images, offline or online (we don't need online, but a lot of the algorithms are developed for real time augmented reality applications).
 
Google: SLAM with camera
* http://answers.ros.org/question/9091/slam-with-one-camera/
 
* [http://openslam.org/ekfmonoslam.html EKFMonoSlam]
* [http://wiki.ros.org/ScaViSLAM ScaViSLAM] - <i>At the moment it supports calibrated stereo rigs, and RGB-D cameras. Monocular SLAM is NOT supported (yet).</i> [https://github.com/strasdat/ScaViSLAM/blob/master/README github]
* [http://www.robots.ox.ac.uk/~mobile/wikisite/pmwiki/pmwiki.php?n=Software.FABMAP FABMAP]
* [http://www.robots.ox.ac.uk/~gk/PTAM/ PTAM] [http://ewokrampage.wordpress.com/ blog] non-commercial GPL (possible??) on [https://github.com/Oxford-PTAM/PTAM-GPL github]. 2013.
* [http://www.doc.ic.ac.uk/~ajd/Scene/ SceneLib] the original? Andrew Davidson & Paul Smith

Latest revision as of 22:46, 16 February 2014

TLV3D

The idea - to map TLV in 3D using any means possible.

We maintain the data on official TAMI superclusters and release it as emperors of the digital realm.

Alon translation: open data map of Tel Aviv (same license as openstreetmaps).

what already exists

Under open licenses (CC / other)

http://openslam.org/ seems to be parking space for slam projects - so far haven't found any full solutions here.

point cloud library http://docs.pointclouds.org/trunk/a02944.html

Other not open licenses

metroscape has 3d models of a few tens of bldgs in tlv

  • requires windows/mac for the plugin, and no download link for any model (STL/other). -- Alon

googleearth has also a few tens of tlv buildings

Googeearth.jpg

will google come and nail it all down ? if we scoop them we get to ride in the google helicopter.

how to

The 3d scanners such as mantis and kinect are apparently not set up for long ranges like 10-50m which is what we're after to get buildings. Satellite photos probably are not of help here due to shallow angles and probable closed source-ness. A quadcopter may be of use but would take massive time and batteries. Rough order of magnitude of the problem:

one street every 50m

about 20km^2 of area in TLV

--> 800 km of street in TLV

--> a car / bike mounted 3d scanner is needed.

photogeometry

from several cameras with known relative angles and baseline, reconstruct 3d info using known algorithms

structured light

try to obtain 3d info from a large-ified kinect hack

it is possible to build a massive structured light rig, using a 3000+ ansi projector and a DSLR.
see here > https://code.google.com/p/structured-light/

15000ansi projector - its 100nis per 1000 lumans per day i tried that method and it works well, but you can't be moving to much. you need to hit the same geometry 3 times with different "binary" patterns.

stereo (camera + laser, use parallex) maker scanner

an opensource diy scanner here

Time of Flight

Cheap LiDARs (relatively) @Hokuyo

  • 10m distance (this model, there are others, maybe 30m? also indoors/outdoors differs)

VL6180 by ST.

Structure from Motion (SfM)

Wikipedia article.

Not necessarily with an IMU, just a camera.

Hardware

Cameras

Map Stitching


Monocular Slam

Monocular is a code word for a single camera, here is a list of algorithms, software has still not been tested.

Input is a bunch of camera images, offline or online (we don't need online, but a lot of the algorithms are developed for real time augmented reality applications).

Google: SLAM with camera