Omnidirectional Vision for Mobile Robotspsfmr.univpm.it/slide/Menegatti.pdfFrom an optical point of view: ~180º FOV wide FOV dioptric cameras (e.g. fisheye) ... – Laser range finder

Omnidirectional Vision for Omnidirectional Vision for Mobile RobotsMobile Robots

Emanuele MenegattiEmanuele Menegatti

IASIAS--Lab Lab Intelligent Autonomous Intelligent Autonomous

Systems LaboratorySystems Laboratory

University of PaduUniversity of Paduaa

ITALYITALY

E. Menegatti - Omnidirectional vision 2

HistoryHistory

Animals Animals PaintingsPaintingsPanoramasPanoramasPhotographyPhotographyComputer VisionComputer VisionRobot VisionRobot Vision

ChapterChapter 1:1:

3

Animal Vision Animal Vision -- 112 Millions species2 Millions speciesFew hundred distinct organisational plansFew hundred distinct organisational plans

BUT BUT only two basic types of only two basic types of imaging eyes of wide use:imaging eyes of wide use:

••Single lens camera like eyesSingle lens camera like eyes

••MultiMulti--lens compound eyeslens compound eyesFrom R. Dawkins: Climbing Mount Improbable. Norton, 1996


AnimalAnimal Vision Vision -- 22Q: Q: Why are perspective systems insufficient and why is field of Why are perspective systems insufficient and why is field of view important?view important?A: A: Perspective systems are one imaging modality of many, we Perspective systems are one imaging modality of many, we are interested in sensors better suited to specific tasks. Sensoare interested in sensors better suited to specific tasks. Sensor r modality should enter into design of computer vision systemsmodality should enter into design of computer vision systems

For example, perhaps for flight For example, perhaps for flight wide fieldwide field--ofof--view sensorsview sensors are are appropriate, and in general useful appropriate, and in general useful for mobile robots.for mobile robots.

[Slide from C. Geyersee Reference Section]


OmnivisionOmnivision in Animalsin Animals

500 million years ago500 million years agoTrilobitesTrilobitesDiurnal InsectsDiurnal InsectsNocturnal InsectsNocturnal InsectsCrustaceousCrustaceousGigantocyprisGigantocypris(not (not omnivisionomnivision))


OmnivisionOmnivision in Paintingsin Paintings

““The wedding of Giovanni The wedding of Giovanni ArnolfiniArnolfini” J. Van ” J. Van EyckEyck 13901390--14411441

Witch MirrorsWitch Mirrors


OmnivisionOmnivision in Paintingsin Paintings

““The praetor and his wife” Q. The praetor and his wife” Q. MetsysMetsys 14661466--15301530Witch MirrorsWitch Mirrors


Distortion in PaintingsDistortion in Paintings

AnamorphosisAnamorphosis


PanoramasPanoramas

Patented by Patented by Robert BarkerRobert Barker -- 17871787

(meaning of the XIX century)(meaning of the XIX century)


PhotographyPhotography

Edinburgh Edinburgh -- Courtesy of Courtesy of EdVecEdVec

MaunaMauna--Kea ObservatoryKea Observatory


Omnidirectional Cameras Omnidirectional Cameras -- 11

CompoundCompound--eye eye camera camera

(from (from Univ. of Maryland, College Park. )

Panoramic cameras Panoramic cameras (from Apple)(from Apple)

Omnidirectional Omnidirectional camerascameras

(from (from University of Picardie - France)


Omnidirectional Cameras Omnidirectional Cameras -- 22Omnidirectional sensors come in many varieties, but by Omnidirectional sensors come in many varieties, but by definition must have a wide fielddefinition must have a wide field--ofof--view. view. From an optical point of view:From an optical point of view:

~180~180ºº FOVFOV

wide FOV wide FOV dioptricdioptriccameras (e.g. fisheye)cameras (e.g. fisheye)

~360~360ºº FOVFOV

polydioptricpolydioptric cameras (e.g. cameras (e.g. multiple overlapping cameras)multiple overlapping cameras)

>180º FOV

catadioptriccatadioptric cameras (e.g. cameras (e.g. cameras and mirror systems)cameras and mirror systems)

[Slide adapted from C. Geyersee Reference Section]

14

CompoundCompound--eye cameraseye cameras

The The RingcamRingcam at at Microsoft ResearchMicrosoft Research

ViewplusViewplusSoftpiaSoftpia Japan & Gifu UniversityJapan & Gifu University

Pros:- High resolutionper viewing angle

Cons:- Bandwidth- Multiple cameras

calibrating and synchronizing;

- expensive

LadybugLadybugPointGreyPointGrey

http://www.ptgrey.com/products/pands.html

15

CompoundCompound--eye cameraseye cameras (inward)(inward)

Virtualised RealityVirtualised RealityCMUCMU

SMART eMotion (Spin-off Univ. of Padua)

now BTS Bioengineering


Panoramic camerasPanoramic cameras

CirkutCirkut CameraCamera

Panning cameraPanning camera Swing lensSwing lens

RoundshotRoundshot

Pros:- High resolutionper viewing angle

Cons:- slow acquisition;- No dynamic scene- expensive

17

Omnidirectional cameras Omnidirectional cameras -- 11

PALPALPanoramic Annular lensPanoramic Annular lens

CatadioprticCatadioprticSensorSensor Two folded mirrorTwo folded mirror

sensorsensorCons:- Blindspot- Low resolution

Pros:- Single image

18

Omnidirectional cameras Omnidirectional cameras -- 22

Cons:- Low resolution at periphery

Pros:- Single image

[Slide adapted from T. Pajdlasee Reference Section]


Q: What kind of sensor should one use?Q: What kind of sensor should one use?A: Depends on your application.A: Depends on your application.1.1. If you are primarily concerned with:If you are primarily concerned with:

–– resolutionresolution –– surveillance (coverage)surveillance (coverage)

and can afford the bandwidth & expense,and can afford the bandwidth & expense,you might stick with you might stick with polydioptricpolydioptric solutionssolutions

2.2. If you are concerned withIf you are concerned with–– bandwidth bandwidth --mobile robots mobile robots ––servoingservoing, SFM, SFM

investigate investigate catadioptriccatadioptric or single wideor single wideFOV FOV dioptricdioptric solutionssolutions

Confused?Confused?Confused?Confused?Confused?Confused?Confused?


http://www.cis.upenn.edu/~kostas/omnigrasp.htmlhttp://www.fxpal.xerox.com/smartspaces/flycam/flycam_home.htm


Other myths and hesitationOther myths and hesitation--11Myth:Myth: CatadioptricCatadioptric images are by necessity highly distorted.images are by necessity highly distorted.Truth:Truth:

Actually no; parabolic mirrors induce no distortion Actually no; parabolic mirrors induce no distortion (perpendicular to the viewing direction).(perpendicular to the viewing direction).With Hyperbolic mirrors is possible to reconstruct With Hyperbolic mirrors is possible to reconstruct perspective imagesperspective images

Myth:Myth: Omnidirectional cameras are more complicated than Omnidirectional cameras are more complicated than perspective cameras, and harder to do SFM with.perspective cameras, and harder to do SFM with.

Truth:Truth: Actually no; parabolic mirrors are easy to model, Actually no; parabolic mirrors are easy to model, calibrate and do SFM with.calibrate and do SFM with.



Other myths and hesitationOther myths and hesitation--22Truth:Truth: Omnidirectional systems have lower resolutionOmnidirectional systems have lower resolution

(because they collect light from a larger FOV and (because they collect light from a larger FOV and capture it with an image sensor with the same size of capture it with an image sensor with the same size of perspective cameras)perspective cameras)

Tradeoff:Tradeoff:Balance resolution and field of view for your needsBalance resolution and field of view for your needsEspecially exploiting the flexibility of Especially exploiting the flexibility of catadioptriccatadioptric systems systems in which by in which by chosingchosing the curvature of the mirror profile the curvature of the mirror profile you can chose the resolution in different portion of the you can chose the resolution in different portion of the imageimage



CatadioptricCatadioptric CameraCamera

Composed of:Composed of:Standard CameraStandard CameraConvex MirrorConvex MirrorSupportSupport–– transparent cylindertransparent cylinder–– Lateral barLateral bar


An omnidirectional camera viewAn omnidirectional camera view

Panoramic Cylinder Panoramic Cylinder ((fromfrom Centre for Machine Perception, Praha)

OmnidirectionalOmnidirectionalimageimage


Omnidirectional camerasOmnidirectional cameras

AdvantagesAdvantagesWide vision fieldWide vision field

OneOne--shot imageshot image

High speedHigh speedVertical LinesVertical LinesRotational InvarianceRotational Invariance

Customisable field of viewCustomisable field of view

Customisable resolutionCustomisable resolution

DisadvantagesDisadvantagesLow ResolutionLow ResolutionDistortionsDistortions

Low readabilityLow readability


Omnidirectional Mirror DesignOmnidirectional Mirror DesignChapter 2:Chapter 2:

The aim is to create a custom mapping between the The aim is to create a custom mapping between the world coordinates and the image coordinates.world coordinates and the image coordinates.

Families of Mirrors Families of Mirrors Geometrical Solid FiguresGeometrical Solid FiguresCustom profileCustom profile

Algorithm for mirror designAlgorithm for mirror designx

y P

26

Families of Mirrors Families of Mirrors -- 11

Geometrical Solid Figures Geometrical Solid Figures (from H. Ishiguro)(from H. Ishiguro)

27

The mirror we designed...The mirror we designed...

Three parts:Three parts:Measurement MirrorMeasurement MirrorMarker MirrorMarker MirrorProximity MirrorProximity Mirror

Mirror ProfileMirror ProfileThe task determines the

mirror profile

E. Menegatti, E. Pagello, et al.Designing an omnidirectional vision system for a goal keeper robotRoboCup-2001: Robot Soccer World Cup V. (Springer 2001)

28

Examples of IASExamples of IAS--Lab MirrorsLab Mirrors

FraunhoferFraunhofer Institute’s MirrorInstitute’s MirrorRoboCup MirrorRoboCup Mirror

Smallest MirrorSmallest MirrorNow, these mirrors are sold by our SpinNow, these mirrors are sold by our Spin--off company off company IT+RoboticsIT+Robotics


IT+Robotics IT+Robotics srlsrlIT+Robotics is a SpinIT+Robotics is a Spin--Off Company of the Off Company of the University of PaduaUniversity of Padua

Its business are:Its business are:••Omnidirectional vision systems (HW & SW)Omnidirectional vision systems (HW & SW)••Intelligent VideoIntelligent Video--surveillance Systemssurveillance Systems••SmallSmall--scale Humanoid Robotsscale Humanoid Robots

Link: Link: http://www.it-robotics.it/

E. Menegatti - Omni Vision 30

Map matchingMap matchingImageImage--based localizationbased localizationObservation of Optical Flow Observation of Optical Flow BiomimeticBiomimetic BehavioursBehavioursIntegration of OmniIntegration of Omni--vision with other sensors:vision with other sensors:–– SonarSonar–– Laser range finderLaser range finder

Outdoor NavigationOutdoor NavigationSLAM (Simultaneous Localization And Mapping)SLAM (Simultaneous Localization And Mapping)Environment reconstruction & 3D mappingEnvironment reconstruction & 3D mappingMiscellaneaMiscellanea

Chapter Chapter 3:3:

OmniOmni--Vision for Mobile RobotsVision for Mobile Robots


Navigation/Localization TricksNavigation/Localization Tricks

Invariance of AzimuthInvariance of AzimuthRotational InvarianceRotational InvarianceVertical Lines mapped in radial linesVertical Lines mapped in radial linesCircumferential continuityCircumferential continuityPeriodicity of the imagePeriodicity of the imageRobustness to occlusionRobustness to occlusion


Invariance of AzimuthInvariance of Azimuth

The azimuth of the object is maintained by the sensor


Rotational InvarianceRotational Invariance

Initial Position Image Counter-Rotated Robot Rotated by 90°

P2P2

P4P4 P3P3

P5P5

P1P1


Vertical Lines Vertical Lines radial linesradial lines

Original Image Edge Detection+

Hough Transform

New Design to stretchvertical lines


Continuity & PeriodicityContinuity & Periodicity

Original Image

Panoramic Cylinder

Fourier Transform


Robustness to occlusionRobustness to occlusionThanks to the wide FOV, usually occluding objects do not change Thanks to the wide FOV, usually occluding objects do not change much the imagemuch the imageSeveral similarity measures have been proved to be robust to Several similarity measures have been proved to be robust to occlusionocclusion

Extreme case presented by Extreme case presented by JoganJogan & & LeonardisLeonardis

Matjaž Jogan, Aleš Leonardis“Robust localization using an “Robust localization using an

omnidirectional appearanceomnidirectional appearance--basedbasedsubspace model of environment”subspace model of environment”

Robotics and Autonomous Systems 45 (2003) 51–72


ApplicationsApplications

Map matchingMap matchingImageImage--based localizationbased localizationObservation of Optical Flow Observation of Optical Flow BiomimeticBiomimetic BehavioursBehavioursIntegration of OmniIntegration of Omni--vision with other sensors:vision with other sensors:–– SonarSonar–– Laser range finderLaser range finder

Outdoor NavigationOutdoor NavigationSLAM (Simultaneous Localization And Mapping)SLAM (Simultaneous Localization And Mapping)

Environment reconstruction & 3D mappingEnvironment reconstruction & 3D mappingMiscellaneaMiscellanea


Map matching Map matching -- 11

YagiYagi used the vertical used the vertical edges of the objects to edges of the objects to find position of the robot find position of the robot on a mapon a mapEdges trackingEdges tracking

Y. Yagi, Y. Nishizawa, M. Yachida,MapMap--Based Navigation for A Mobile Robot with Based Navigation for A Mobile Robot with OmnidirectionalOmnidirectional Image SensorImage Sensor COPIS,COPIS,

IEEE Trans. Robotics and Automation,pp.634-648,Vol.11,No.5,1995.10


Map matching Map matching -- 22

Menegatti et al. used the Menegatti et al. used the Chromatic Transitions of Chromatic Transitions of Interest to perform scan Interest to perform scan matchingmatchingMonteMonte--Carlo Localization Carlo Localization AlgorithmAlgorithmAlmost the same Almost the same approach used with Laser approach used with Laser range Findersrange Finders

E. Menegatti, A. E. Menegatti, A. PrettoPretto, A. , A. ScarpaScarpa, E. Pagello, E. PagelloOmnidirectional vision scan matching for robot localization in dOmnidirectional vision scan matching for robot localization in dynamic environments ynamic environments

IEEE Transactions on Robotics, Vol. 22,IEEE Transactions on Robotics, Vol. 22, Iss. 3  June 2006  pages 523Iss. 3  June 2006  pages 523-- 535 535 


ImageImage--based navigation based navigation -- 11Ishiguro and Menegatti:Ishiguro and Menegatti:–– FFT magnitude for positionFFT magnitude for position–– FFT phase for headingFFT phase for heading–– SelfSelf--organization of the memoryorganization of the memory–– ImageImage--based Localisationbased Localisation–– Hierarchical LocalizationHierarchical Localization–– ImageImage--Based Monte Carlo LocalisationBased Monte Carlo Localisation

Emanuele Menegatti, M. Zoccarato, E. Pagello, H.Ishiguro, ````ImageImage--based Montebased Monte--Carlo Localisation with Carlo Localisation with Omnidirectional images'' Omnidirectional images''

Robotics and Autonomous Systems, Elsevier - 2004

Emanuele Menegatti, Takashi Maeda, Hiroshi Ishiguro, ````Hierarchycal ImageHierarchycal Image--based Memory for Robot Navigationbased Memory for Robot Navigation,'' ,''

Robotics and Autonomous Systems, Elsevier - 2004


ImageImage--based navigation based navigation -- 22KröseKröse et al:et al:–– Used Principal Component Analysis to Used Principal Component Analysis to

extract linear featureextract linear feature–– Dataset described in term of Dataset described in term of

eigenimageseigenimages–– Probabilistic localizationProbabilistic localization

B. Kröse, N. Vlassis, R. Bunschoten, and Y. Motomura. “A probabilistic model for “A probabilistic model for appareanceappareance--based based

robot localization”robot localization”Image and Vision Comp, vol. 19(6):pp. 381–391, April

2001..


ImageImage--based navigation based navigation -- 33Gross et al:Gross et al:–– Used slices of the Used slices of the

panoramic cylinderpanoramic cylinder–– Slices confronted via Slices confronted via

colour histogramscolour histograms–– Hybrid map: Hybrid map:

topological map topological map aumentedaumented with with metric information metric information

T. Wilhelm, H.T. Wilhelm, H.--J. J. BöhmeBöhme, and H., and H.--M. Gross. M. Gross. ““A multiA multi--modal system for tracking and analyzing faces on a mobile robot”modal system for tracking and analyzing faces on a mobile robot”

Robotics and Autonomous Systems, 48:31Robotics and Autonomous Systems, 48:31––40, August 2004.40, August 2004.


Observation of Optical FlowObservation of Optical Flow

Hiroshi Ishiguro, Kenji Ueda and Saburo Tsuji, ``Omnidirectional Visual Information``Omnidirectional Visual Information forfor NavigatingNavigating a Mobile Robota Mobile Robot'', '', IEEE Int. Conf.

on Robotics and Automation (ICRA-93), pp. 799-804, 1993.

Ishiguro used:Ishiguro used:–– Foci of Expansion (FOE) to Foci of Expansion (FOE) to

estimate relative positionsestimate relative positions–– No encoder infoNo encoder info

Svoboda used:Svoboda used:–– Optical flow to discriminate Optical flow to discriminate

translation and rotationstranslation and rotations

Tomáš Svoboda, Tomáš Pajdla, and Václav Hlavác.“Motion estimation using central panoramic cameras”

IEEE Int. Conf. on Intelligent Vehicles, 1998.


BiomimeticBiomimetic BehavioursBehavioursArgyros, A.A.; Tsakiris, D.P.; Groyer, C.

Biomimetic centering behavior Robotics & Automation Magazine, IEEE Publication Date: Dec. 2004 . Vol.11,

Iss. 4 pp.21- 30

M.V. Srinivasan. A new class of mirrors for wide-angle imaging.

Proceedings, IEEE Workshop on Omnidirectional Vision and Camera Networks. Madison, Wisconsin, USA., June 2003.

G.L. Barrows, J.S. Chahl and M.V. Srinivasan (2003) Biomimetic visual sensing and flight control. The Aeronautical Journal, London: The Royal Aeronautical

Society, vol, 107, No. 1069, pp. 159-168.


Integration with other sensorsIntegration with other sensors

Shin-Chieh Wei, Yasushi Yagi and Masahiko Yachida,“On“On--line Map Building Based On Ultrasonic and Image Sensor, 1996 line Map Building Based On Ultrasonic and Image Sensor, 1996 IEEE Int. Conf.

on Robotics and Automation(ICRA-98) 1998

YagiYagi used:used:–– Sonar to detect free spaceSonar to detect free space–– Fused the sonar, edge, Fused the sonar, edge,

colour information in a colour information in a occupancy grid occupancy grid

ClerentinClerentin used:used:–– Laser to find rangeLaser to find range–– Fused laser and edgesFused laser and edges

A. Clerentin, L. Delahoche, C. Pegard, E. Brassart"A localization method based on two A localization method based on two omnidirectionalomnidirectional perception systems cooperationperception systems cooperation “

ICRA'2000, San Francisco, April 2000.


Outdoor Navigation Outdoor Navigation -- 11

OmnidirectionalOmnidirectional Vision for Vision for Road Following with NN:Road Following with NN:–– Road classification Road classification –– Steering angleSteering angle

ZZ.. Zhu, SZhu, S. . Yang, GYang, G. . XuXu, X, X..Lin, Lin, DingjiDingji Shi Shi "Fast road classification and orientation estimation using omni"Fast road classification and orientation estimation using omni--view images and view images and

neural networks," neural networks," IEEE Transaction on Image Processing, Vol. 7, No.8, August 1998, pp. 1182-1197.



Paul Paul BlaerBlaer and Peter Allenand Peter Allen“Topological Mobile Robot Localization Using Fast Vision Techniq“Topological Mobile Robot Localization Using Fast Vision Techniques”ues”

Proceedings of the 2002 IEEE International Conference on RoboticProceedings of the 2002 IEEE International Conference on Robotics & Automation s & Automation 20022002

ImageImage--based navigation:based navigation:–– Topological navigationTopological navigation–– Histogram matchingHistogram matching–– Different localisation Different localisation

accuraciesaccuracies



JoséJosé--Joel GonzalezJoel Gonzalez--BarbosaBarbosa and Simon and Simon LacroixLacroixRover localization in natural environments by indexing panoramicRover localization in natural environments by indexing panoramic imagesimages

Proceedings of the 2002 IEEE International Conference on RoboticProceedings of the 2002 IEEE International Conference on Robotics & s & Automation 2002Automation 2002

ImageImage--based navigation:based navigation:–– Dimension reduction with Dimension reduction with

PCAPCA–– Histogram matchingHistogram matching


SLAMSLAMMichael Michael KaessKaess and Frank and Frank DellaertDellaert, , 

Visual SLAM with a MultiVisual SLAM with a Multi--Camera RigCamera Rig, , Georgia Tech Technical Report GITGeorgia Tech Technical Report GIT--GVUGVU--

0606--06, 200606, 2006

Thomas Thomas LemaireLemaire, Simon , Simon LacroixLacroix. . Long Term SLAM with panoramic vision. Long Term SLAM with panoramic vision.

Submitted to Journal of Fields RoboticsSubmitted to Journal of Fields Roboticsspecial issue on special issue on

"SLAM in the Fields"."SLAM in the Fields".


Environment ReconstructionEnvironment Reconstruction

H.BaksteinH.Bakstein, , T.PajdlaT.Pajdla. Rendering Novel Views from a Set . Rendering Novel Views from a Set of Omnidirectional Mosaic Images. Workshop onof Omnidirectional Mosaic Images. Workshop on

Omnidirectional Vision and Camera Networks 2003, CD Omnidirectional Vision and Camera Networks 2003, CD ROM, IEEE June 2003.ROM, IEEE June 2003.

C. Geyer, K. C. Geyer, K. DaniilidisDaniilidis,,Mirrors in Motion: Mirrors in Motion: EpipolarEpipolar geometry and motion geometry and motion

estimation, estimation, c. Inter. Conf. on Computer Vision, October, 2003, c. Inter. Conf. on Computer Vision, October, 2003,

Nice, France.Nice, France.


Omnidirectional Distributed Vision System (ODVS)

Requirements:•Robots’ only sensor: omnidirectional vision•No use of external computer•Every robot shares its measures•Every robot fuses all measures received by teammates•Measures can refer to different instants in time

The aim of the ODVS: to track multiple moving objects in highly dynamic environments by sharing the information gathered by every single robot

E. Pagello, A. E. Pagello, A. D’AngeloD’Angelo, E. Menegatti , E. Menegatti Cooperation Issues and Distributed Sensing for MultiCooperation Issues and Distributed Sensing for Multi--Robot Systems Robot Systems IEEE Proceedings of IEEE (in press due October 2006)IEEE Proceedings of IEEE (in press due October 2006)


OneOne StaticStatic Vision Vision AgentAgent ((omnidirectionalomnidirectional camera)camera)

FiveFive StaticStatic AcusticAcustic AgentsAgents ((steerablesteerable microphonemicrophonearraysarrays))

OneOne Mobile Vision Mobile Vision AgentAgent (robot (robot withwith omnidirectionalomnidirectionalcamera)camera)

=

=

Experiment Layout

Audio-Video Surveillance System with Mobile Robot

E. Menegatti, M. E. Menegatti, M. CavasinCavasin, E. Mumolo, M. , E. Mumolo, M. NolichNolich, E. Pagello , E. Pagello CombiningCombining Audio and Video Audio and Video SurveillanceSurveillance withwith a Mobile Robot a Mobile Robot InternationalInternational Journal on Journal on ArtificialArtificial Intelligence Intelligence ToolsTools (in press)(in press)

=


The End!The End!

Thanks for your attention!

My publications and more information can be found in my web page:

http://www.dei.unipd.it/~emg


ReferencesReferencesWWW:The page of omnidirectional visionhttp://www.cis.upenn.edu/~kostas/omni.html

ICCV03 Course on Omnidirectional Visionhttp://www.cis.upenn.edu/~kostas/omni/iccv03.html

Book:R. Benosman & S.B. Kang (Eds.)Panoramic VisionSpringer 2001

http://www.cis.upenn.edu/~kostas/omni.htmlhttp://www.cis.upenn.edu/~kostas/omni/iccv03.html


ReferencesReferencesSpecial issues:K. Daniilidis& N. PapanikolopoulosThe Big PictureIEEE Robotics & Automation Magazine Dec. 2004

Hiroshi Ishiguro and Ryad BenosmanSpecial issue on omnidirectional vision and its applications Machine Vision and Applications (2003) Vol 14

Yasushi Yagi and Katsushi IkeuchiSpecial Issue on Omni-Directional Research in JapanInternational Journal of Computer Vision Vol. 58, Num. 3, Springer July 2004

Peter Sturm, Tomas Svoboda and Seth TellerSpecial issue on Omnidirectional Vision and Camera NetworksComputer Vision and Image Understanding Vol. 103, Iss. 3, Sept. 2006

Omnidirectional Vision for Mobile RobotsHistoryAnimal Vision - 1Animal Vision - 2Omnivision in AnimalsOmnivision in PaintingsOmnivision in PaintingsDistortion in PaintingsDistortion in PaintingsPanoramasPhotographyOmnidirectional Cameras - 1Omnidirectional Cameras - 2Compound-eye camerasCompound-eye cameras (inward)Panoramic camerasOmnidirectional cameras - 1Omnidirectional cameras - 2Confused?Other myths and hesitation-1Other myths and hesitation-2Catadioptric CameraAn omnidirectional camera viewOmnidirectional camerasOmnidirectional Mirror DesignFamilies of Mirrors - 1The mirror we designed...Examples of IAS-Lab MirrorsIT+Robotics srlOmni-Vision for Mobile RobotsNavigation/Localization TricksInvariance of AzimuthRotational InvarianceVertical Lines radial linesContinuity & PeriodicityRobustness to occlusionApplicationsMap matching - 1Map matching - 2Image-based navigation - 1Image-based navigation - 2Image-based navigation - 3Observation of Optical FlowBiomimetic BehavioursIntegration with other sensorsOutdoor Navigation - 1Outdoor Navigation - 2Outdoor Navigation - 3SLAMEnvironment ReconstructionThe End!ReferencesReferences

Documents

Omnidirectional Vision for Mobile Robotspsfmr.univpm.it/slide/Menegatti.pdfFrom an optical point of view: ~180º FOV wide FOV dioptric cameras (e.g. fisheye) ... – Laser range finder