| 期刊名称： |
Remote Sensing |
| 全部作者： |
Bo Guo,Xianfeng Huang,Qingquan Li,Fan Zhang,Jiasong Zhu*,Chisheng Wang |
| 出版年份： |
2016 |
| 卷 号： |
8 |
| 期 号： |
3 |
| 页 码： |
243 |
| 查看全本： |
|
Objectdetectionandreconstructionfromremotelysenseddataareactiveresearchtopicinphotogrammetricandremotesensingcommunities.Powerengineeringdevicemonitoringbydetectingkeyobjectsisimportantforpowersafety.Inthispaper,weintroduceanovelmethodforthereconstructionofself-supportingpylonswidelyusedinhighvoltagepower-linesystemsfromairborneLiDARdata.Ourworkconstructspylonsfromalibraryof3Dparametricmodels,whicharerepresentedusingpolyhedronsbasedonstochasticgeometry.Firstly,laserpointsofpylonsareextractedfromthedatasetusinganautomaticclassificationmethod.Anenergyfunctionmadeupoftwotermsisthendefined:thefirsttermmeasurestheadequacyoftheobjectswithrespecttothedata,andthesecondtermhastheabilitytofavororpenalizecertainconfigurationsbasedonpriorknowledge.Finally,estimationisundertakenbyminimizingtheenergyusingsimulatedannealing.WeuseaMarkovChainMonteCarlosampler,leadingtoanoptimalconfigurationofobjects.Twomaincontributionsofthispaperare:(1)buildingaframeworkforautomaticpylonreconstruction;and(2)efficientglobaloptimization.Thepylonscanbepreciselyreconstructedthroughenergyoptimization.Experimentsproducingconvincingresultsvalidatedtheproposedmethodusingadatasetofcomplexstructure.
