Structure from motion: differenze tra le versioni

Contenuto cancellato Contenuto aggiunto
Recupero di 1 fonte/i e segnalazione di 0 link interrotto/i.) #IABot (v2.0.9.5
 
(42 versioni intermedie di 23 utenti non mostrate)
Riga 1:
[[File:DSM_construction_siteDSM construction site.jpg|thumb|right| Modello di superficie digitale di un sito di costruzione di una [[Intersezione a livelli sfalsati]] di un'autostrada ]]
'''Structure from motion''' o '''SfM''' (in italiano: ''Struttura dal movimento'') è una tecnica di [[range imaging]] della [[computer vision]] e della [[Vista (senso)|percezione visiva]], con cui il processo di stima di strutture tridimensionali da sequenze di immagini bidimensionalbidimensionali che può essere accoppiato con segnali di movimento locali.
Nella visione biologica la SfM fa riferimento al fenomeno per cui gli esseri umani ( e altre creature viventi) posso ricostruire strutture 3D da un campo in movimento in proiezione 2D (retinale) di un oggetto o una scena in movimento.
 
== Ottenere informazione 3D da immagini 2D ==
Gli esseri umani ricevono molte informazioni sulla struttura tridimensionale nel loro ambiente muovendosi attraverso.
Quando l'osservatore si muove e gli oggetti intorno a lui si muovono, l'informazione è ottenuta da immagini nel tempo<ref>{{cite bookcita libro| title titolo= Computer Vision |url=https://archive.org/details/computervision0000shap author |autore= Linda G. Shapiro, George C. Stockman | publisher editore= Prentice Hall | year anno= 2001 | isbn = 0-13-030796-3 }}</ref>.
Trovare strutture dal movimento presenti problemi simili a trovare strutture dalla visione stereoscopica. In entrambi i casi, la corrispondenza tra immagini e la ricostruzione di ogettioggetti 3D ha bisogno di essere trovata.
 
Per trovare corrispondenza tra immagini, le caratteristiche (feature) come i punti d'angolo (lati con gradienti in direzioni multiple) sono tracciate da un 'immagine alla prossima.
UnaUno delle piùdei rilevatori di caratteristiche più usati è il [[Scale-invariant feature transform|SIFT]] (Scale-invariant feature transform). Usa come caratteristiche il massimo di una piramide di [[differenzaDifferenze di gaussianeGaussiane]] ([[DOG]])
Il primo passo nella SIFT è di trovare una direzione di gradiente dominante. Per renderla invariante alla rotazione, il [[Descrittore (informazione)|descrittore]] è ruotato per adattarne l'orientamento<ref>{{Cite journalCita pubblicazione| journal rivista= International Journal of Computer Vision | title titolo= Distinctive image features from scale-invariant keypoints | author autore= D. G. Lowe | year anno= 2004 }}</ref> .
Un' altro rilevatore di caratteristiche molto comune è il [[SURF]] ([[Speeded Up Robust Features]]). Nel SURF, la DOG è sostituita da un rilevatore hessiano di [[Riconoscimento di regioni|blob]] basato su matrici<ref>{{Cite journalCita pubblicazione| journal rivista= 9th European Conference on Computer Vision | title titolo= Surf: Speeded up robust features | author autore= H. Bay, T. Tuytelaars, and L. Van Gool | year anno= 2006 }}</ref>.
Inoltre, invece di valutare il gradiente degli istogrammi, SURF calcola le somme dei componenti del gradiente e le somme dei loro valori assoluti<ref>{{Cite journalCita pubblicazione| journal rivista= Kybernetika | title titolo= The structure-from-motion reconstruction pipeline – a survey with focus on short image sequences | author autore= K. Häming and G. Peters| year anno= 2010 | url = http://dml.cz/dmlcz/141400 }}</ref>.
Le caratteristiche rilevate da tutte le immagini saranno poi incrociate.
Uno degli algoritmi di matching (per incrociare i dati) che traccia le caratteristiche da un'immagine all'altra è il [[tracciatore Lukas-Kanade]]<ref>{{Cite journalCita pubblicazione| journal rivista= IJCAI81 | title titolo= An iterative image registration technique with an application to stereo vision | author autore= B. D. Lucas and T. Kanade }}</ref>.
 
== SfM per l'analisi di struttura del patrimonio culturale ==
{{...|informatica}}
 
== Software Structure from Motion ==
{| class="wikitable sortable"
|+Soluzioni Open source
!Nome
!Licenza
!GUI
!Linguaggio
!Descrizione
|-
|[http://phototour.cs.washington.edu/bundler/ Bundler]
|[[GNU General Public License|GNU GPL v2]]
|no
|C++
|Structure from Motion for Unordered Photo Collections by Noah Snavely
|-
|[https://github.com/openMVG/openMVG openMVG]
|[[MPL 2.0|MPL v2]]
|no
|C++
|An Open Multiple View Geometry library + Structure from Motion demonstrators
|-
|[https://developer.blender.org/project/profile/59/ LibMV]
|[[Licenza MIT|MIT]]
|no
|C++
|Library for Multiview Reconstruction (LMV), libreria alla base degli algoritmi di SfM di [[Blender (programma)|Blender]]
|-
|[https://web.archive.org/web/20151223064447/http://cs.ucsb.edu/~cmsweeney/theia/ Theia]
|[[Licenze BSD|BSD]]
|no
|C++
|A Fast and scalable structure-from-motion library
|-
|[http://logiciels.ign.fr/?Telechargement,20 MicMac]
|[[CeCILL|CeCILL-B]]
|si
|C++
|Programma di fotogrammetria de l'[[Institut national de l'information géographique et forestière]]
|-
|[https://web.archive.org/web/20141014235855/http://www.gris.informatik.tu-darmstadt.de/projects/multiview-environment/ MVE]
|[[GNU General Public License|GNU GPL v3]]
|si
|C++
|The Multi-View Environment
|-
|[https://colmap.github.io/ Colmap]
|[[Licenze BSD|BSD]]
|si
|C++
|Pipeline Structure-from-Motion (SfM) e Multi-View Stereo (MVS) a linea di comando e con interfaccia grafica.
|-
|[https://alicevision.github.io/ AliceVision]
|[[MPL 2.0|MPL v2]]
|no
|C++
|Libreria software per la fotogrammetria
|-
|[http://vision.ucsd.edu/~vrabaud/toolbox/doc/overview.html Structure from Motion toolbox for Matlab] {{Webarchive|url=https://web.archive.org/web/20080524044956/http://vision.ucsd.edu/~vrabaud/toolbox/doc/overview.html |date=24 maggio 2008 }}
|[[GNU General Public License|GNU GPL v3]]
|si
|Matlab
|Toolbox creato da Vincent Rabaud
|-
|[http://www.robots.ox.ac.uk/~vgg/hzbook/code/ Matlab Functions for Multiple View Geometry]
|[[Licenza MIT|MIT]]
|si
|Matlab
|Toolbox creato da Andrew Zissermann
|-
|[http://www.robots.ox.ac.uk/~vgg/hzbook/code/ Structure and Motion Toolkit]
|MSR-EULA
|si
|Matlab
|Toolbox creato da Phil Torr
|-
|[http://www.cs.dartmouth.edu/~lorenzo/projects/learning-nr-shape/em-sfm.zip Matlab Code for Non-Rigid Structure from Motion]
|n.a.
|si
|Matlab
|Toolbox creato da Lorenzo Torresani
|-
|[[AliceVision Meshroom]]
|[[MPL 2.0|MPL v2]]
|si
|Python
|Applicazione GUI ([[Qt (toolkit)|Qt]]/PySide) basata sulla libreria AliceVision
|}
 
=== Altri software ===
*[http://www.acute3d.com/ Smart3DCapture], a complete photogrammetry solution by Acute3D.
*[http://www.3dflow.net/technology/samantha-structure-from-motion/ 3DF Samantha - Command line structure from Motion pipeline for Windows], by 3Dflow srl. Free for non-commercial purposes.
*[http://www.zjucvg.net/acts/acts.html Automatic Camera Tracking System (ACTS)], a structure-from-motion with dense depth recovery system for Microsoft Windows, by Vision Group of State Key Lab of CAD&CG, Zhejiang University.
*[http://www.zjucvg.net/ls-acts/ls-acts.html Large-Scale Automatic Camera Tracking System (LS-ACTS)], a large-scale structure-from-motion system for Microsoft Windows, by Vision Group of State Key Lab of CAD&CG, Zhejiang University.
*[http://ccwu.me/vsfm VisualSFM: A Visual Structure from Motion System], by Changchang Wu
*{{cita web|http://www.visual-experiments.com/demos/sfmtoolkit/|SFMToolkit a complete photogrammetry solution based on open-source software}}
*[https://web.archive.org/web/20160131001254/http://www.digitalsurf.fr/en/mntsem.html MountainsMap SEM] software for [[Scanning Electron Microscope]]s. 3D is obtained by tilting the specimen + photogrammetry.
*[http://www.viscoda.com/index.php/en/products/non-commercial/voodoo-camera-tracker Voodoo Camera Tracker] {{Webarchive|url=https://web.archive.org/web/20160430142226/http://www.viscoda.com/index.php/en/products/non-commercial/voodoo-camera-tracker |date=30 aprile 2016 }}, non-commercial tool for the integration of virtual and real scenes.<br />Original site, archived: [https://web.archive.org/web/20120423112957/http://www.digilab.uni-hannover.de/docs/manual.html Laboratorium für Informationstechnologie, University of Hannover]
*[https://web.archive.org/web/20140224230522/http://dev.metaio.com/sdk/toolbox/ MetaIO Toolbox] SfM for augmented reality on mobile devices.
*[http://www.2d3sensing.com/content/tacitview TacitView] {{Webarchive|url=https://web.archive.org/web/20150402093429/http://www.2d3sensing.com/content/tacitview |date=2 aprile 2015 }} by [[2d3|2d3 Sensing]]
*[https://catena.googlecode.com/ Catena] {{Webarchive|url=https://archive.is/20140402020553/http://catena.googlecode.com/# |data=2 aprile 2014 }} Python Abstract Workflow Framework with SfM components.
*[https://web.archive.org/web/20160304130300/http://ptak.felk.cvut.cz/sfmservice/websfm.pl?menu=cmpmvs CMPMVS] Multi-View Reconstruction Software
 
== Note ==
<references/>
 
== Bibliografia ==
* {{cita libro|autore=Richard Hartley and Andrew Zisserman |titolo=Multiple View Geometry in Computer Vision |url=https://archive.org/details/multipleviewgeom0000hart |editore=Cambridge University Press|anno=2003 |isbn=0-521-54051-8}}
* {{cita libro|autore=Olivier Faugeras and Quang-Tuan Luong and Theodore Papadopoulo |titolo=The Geometry of Multiple Images |editore=MIT Press|anno=2001 |isbn=0-262-06220-8}}
*{{cita libro|titolo=An Invitation to 3-D Vision: From Images to Geometric Models |anno=2003 |url=https://archive.org/details/invitationto3dvi0000unse |autore=Yi Ma, S. Shankar Sastry, Jana Kosecka, Stefano Soatto, Jana Kosecka |editore=Springer-Verlag New York, LLC |data=novembre 2003|isbn=0-387-00893-4|trasmissione=Interdisciplinary Applied Mathematics Series, #26}}
 
== Voci correlate ==
 
*[[Match moving]]
*[[Motion field]]
*[[Geometria epipolare]]
*[[SLAM (robotica)|SLAM]]
 
== Bibliografia ==
 
* {{cite book |
author=Richard Hartley and Andrew Zisserman |
title=Multiple View Geometry in Computer Vision |
publisher=Cambridge University Press|
year=2003 |
isbn=0-521-54051-8}}
 
* {{cite book |
author=Olivier Faugeras and Quang-Tuan Luong and Theodore Papadopoulo |
title=The Geometry of Multiple Images |
publisher=MIT Press|
year=2001 |
isbn=0-262-06220-8}}
*{{cite book |
title=An Invitation to 3-D Vision: From Images to Geometric Models |
author=Yi Ma, S. Shankar Sastry, Jana Kosecka, Stefano Soatto, Jana Kosecka |
publisher=Springer-Verlag New York, LLC |date=November 2003|
isbn=0-387-00893-4|
series=Interdisciplinary Applied Mathematics Series, #26
}}
 
== Collegamenti esterni ==
*{{cita web|https://exporttocanoma.blogspot.it/2012/11/sfm-structure-from-motion-prima-parte.html|Structure from motion in exporttocanoma.blogspot.it}}
===Software Structure from Motion ===
 
{{portale|informatica}}
====Soluzioni Open source ====
[[Categoria:Elaborazione digitale delle immagini]]
 
C++
*[http://phototour.cs.washington.edu/bundler/ Bundler - Structure from Motion for Unordered Photo Collections] by Noah Snavely
*[https://github.com/openMVG/openMVG openMVG] An Open Multiple View Geometry library + Structure from Motion demonstrators
*[https://developer.blender.org/project/profile/59/ Libmv - A Structure from Motion library]
*[http://cs.ucsb.edu/~cmsweeney/theia Theia]: A Fast and scalable structure-from-motion library released under the BSD license
*[http://logiciels.ign.fr/?Telechargement,20 MicMac, a SFM open-source code released] by the [[Institut national de l'information géographique et forestière]]
*[https://vision.in.tum.de/lsdslam LSD-SLAM]: Large-Scale Direct Monocular SLAM in real-time, by Jakob Engel
*[http://www.gris.informatik.tu-darmstadt.de/projects/multiview-environment/ MVE - The Multi-View Environment] by Simon Fuhrmann, TU Darmstadt.
*[http://ceres-solver.org/ ceres-solver for general non-linear least squares]. Has features for bundle adjustment. Previously used by Google internally for google maps. Released to the public in 2012.
*[https://www.ics.forth.gr/~lourakis/sba SBA] for generic bundle adjustment by Manolis Lourakis.
*[https://github.com/pyp22/NuxSFM-2.0 NuxSFM] NuxSFM 2.0 pacman X86_64 build scripts.
 
Matlab
*[http://vision.ucsd.edu/~vrabaud/toolbox/doc/overview.html Structure from Motion toolbox for Matlab] by Vincent Rabaud
*[http://www.robots.ox.ac.uk/~vgg/hzbook/code/ Matlab Functions for Multiple View Geometry] by Andrew Zissermann
*[http://cms.brookes.ac.uk/staff/PhilipTorr/Code/code_page_4.htm Structure and Motion Toolkit] by Phil Torr
*[http://www.cs.dartmouth.edu/~lorenzo/projects/learning-nr-shape/em-sfm.zip Matlab Code for Non-Rigid Structure from Motion] by Lorenzo Torresani
 
Python
*[http://www.arc-team.homelinux.com/arcteam/ppt.php Python Photogrammetry Toolbox GUI] - an open-source SFM GUI (Easy SfM and dense point cloud estimation launcher) by Pierre Moulon and Arc-Team
 
====Altri software====
*[http://www.acute3d.com/ Smart3DCapture], a complete photogrammetry solution by Acute3D.
*[http://www.3dflow.net/technology/samantha-structure-from-motion/ 3DF Samantha - Command line structure from Motion pipeline for Windows], by 3Dflow srl. Free for non-commercial purposes.
*[http://www.zjucvg.net/acts/acts.html Automatic Camera Tracking System (ACTS)], a structure-from-motion with dense depth recovery system for Microsoft Windows, by Vision Group of State Key Lab of CAD&CG, Zhejiang University.
*[http://www.zjucvg.net/ls-acts/ls-acts.html Large-Scale Automatic Camera Tracking System (LS-ACTS)], a large-scale structure-from-motion system for Microsoft Windows, by Vision Group of State Key Lab of CAD&CG, Zhejiang University.
*[http://ccwu.me/vsfm VisualSFM: A Visual Structure from Motion System], by Changchang Wu
*[http://www.visual-experiments.com/demos/sfmtoolkit/ SFMToolkit a complete photogrammetry solution based on open-source software]
*[http://www.digitalsurf.fr/en/mntsem.html MountainsMap SEM] software for [[Scanning Electron Microscope]]s. 3D is obtained by tilting the specimen + photogrammetry.
*[http://www.viscoda.com/index.php/en/products/non-commercial/voodoo-camera-tracker Voodoo Camera Tracker], non-commerial tool for the integration of virtual and real scenes. <br />Original site, archived: [http://web.archive.org/web/20120423112957/http://www.digilab.uni-hannover.de/docs/manual.html Laboratorium für Informationstechnologie, University of Hannover]
*[https://dev.metaio.com/sdk/toolbox MetaIO Toolbox] SfM for augmented reality on mobile devices.
*[http://www.2d3sensing.com/content/tacitview TacitView] by [[2d3|2d3 Sensing]]
*[http://catena.googlecode.com Catena] Python Abstract Workflow Framework with SfM components.
*[http://ptak.felk.cvut.cz/sfmservice/websfm.pl?menu=cmpmvs CMPMVS] Multi-View Reconstruction Software