基于魚眼鏡頭的全方位機(jī)器視覺應(yīng)用研究.pdf

1、天津理工大學(xué)碩士學(xué)位論文基于魚眼鏡頭的全方位機(jī)器視覺應(yīng)用研究姓名：劉麗群申請學(xué)位級別：碩士專業(yè)：機(jī)械電子工程指導(dǎo)教師：曹作良20090101AbstractVisionbasedmobilerobotnavigationiSoneofthehottopicsinrobotfieldrecentlyJustlikehumans，visionisalsothemostimportantsourceforrobottoknowtheworld

2、arounditcansupplytheinformationoftheenvironment，suchastargetidentification、real—timetrackingandnavigationOmni—directionalvisionisanewcomputervisiontechnologywhichcanacquire180一degreesphericalfieldofviewofanenvironmentWim

3、thepropertyoflargefieldofviewOmnidirectionalvisionsystemcallmeetrequirementsthatmachinevisionacquiresteal—timeenvironmentalinformationItisquitesuitableforawiderangeofapplicationsOmni—directionalimagesgeneratedfromomni—di

4、rectionvisionsystemhaveproblemofseriousdistortionbecauseoftheuniqueimagingprocessesofomnidirectionsystem，SOtheinformationofthescenetakenbytheomrlidirectionalvisionneedstoberecoveredwhenapplyingthesystemtopracticesThispap

5、erstudiesthedistortioncorrectionproblemofomni—directionvisionimagetakenbyfisheyelensThisprojectisbasedontheintemationalscienceandtechnologycollaborationprojectsupportedbytheMinistryofScienceandTechnologyofthePeople’SRepu

6、blicofChinaFirstlythetechnologyofomnidirectionalvisionisintroduced，andthebackgroundoftheresearchisdescribed，includingthetheoryandtheapplicationsSecondlythedifferencebetweentheconventionalcamerasandomnidirectionalvisionim

7、agingsystemisdeeplystudied，andthestructureoffisheyecameraandprinciplesofopticalimagingaredetailedBasedonthefisheyecameraimagingmodel，amathematicalmodelisestablishedforthefisheyecameraForcalibrationproblem，thispaperpropos

8、esamethodtogettheinternalparametersoffisheyecameraCompared、析tllthetraditionalmethodofcalibrationthismethodiseasytoapplyandavoidscomplexcalculationTheresultishi【ghpreciseThirdlyaccordingtOthecharacteristicsofthefisheyeima

9、gedistortiontwodistortioncorrectionalgorithmsarederivedinthispaperOnemethodisbasedontheisometricimagemodeloffisheyelensThefunctionoftherelationshipbetweendistortionimageandidealimageisestablishedAnotherisregressionalgori

10、thmbasedonSupportVectorMachine(SVM)whichisnovelWthstrongabilityofclassificationandregression，SVMcallprovideamappingbetweenthefisheyeimageandthestandardimageforhumaneyes，whichinvolvesacoordinateconversionbetweenfisheyeima

11、geandrealworldviewAspecifictrainingtargetisneededforSVMtegressTwomethodsareadoptedbytheVCprogramTheresultsofexperimentsverifythecorrectnessofthealgorithm。Lastlyusinghigh—speedDigitalSignalProcessor(DSP)asthecorehardware，

12、thispapertransplantdistortioncorrectionalgorithmtoDSPcompletingalargenumberofimageprocessinginthelimitedtimeandmeetingtherequirementsofreal—timeThisworklaythefoundationforvisionnavigationandappearsextensiveapplication’Ke