Kristian Hjorth_自动化海底测绘和物体检测.pdf

1、AutomatingAutomating seabedseabed mappingmapping and and objectobjectdetectiondetectionOceanology International,March 2024Kristian Hjorth,Product Manager Argeo SCOPESENSOR TECHNOLOGYARGEO LISTEN +ARGEO WHISPERAUVWROVUSVAll assets are to have Argeo sensor technology embedded Decision ready data to cl

2、ients during mission and project lifecycleRapid data transfer from ship to shore via Argeo SCOPEDIGITAL PLATFORMARGEO SCOPEDATAFrom acquisition to actionable dataCompletesubsea service providerInformation flow during seabed surveysHugin Superior1010 km2Utsira OWF Area:45 Tb9 daysArea surveyed per ho

3、urData collected per day5 Tb4.5 km2APIDIGITAL PLATFORMARGEO SCOPEDecision ready Decision ready data to clients data to clients during mission and during mission and project lifecycleproject lifecycle48 hrsWeeksManual InterpretationObject Detection:Manual WorkflowExample requirements from clients:Det

4、ect and identify all objects larger than 0.5mCross-correlate datasets:Bathymetry BackscatterSAS/SSSMagnetic anomaliesSBPDimensions:Width,length,height(!),burial depthSSDM classifications1010 km2Utsira OWF Area:1.5 M256x256 images:10cm resolutionShadowhAUVObject Detection:Manual WorkflowInternal Expe

5、rience:150km2:3000 contacts(SAS).Estimate:100hrs manual labor per data typeNote:We collect 100km2 per 24hrs.Ten noorden van de WaddeneilandenOWF public data:120km2:2000 contacts(SSS+Mag).Case study:Utsira OWF AreaArgeo collected MC data from Utsira Nord in 2022.Data sources:MBES Bathymetry 50 x50cmM

6、BES Backscatter 50 x50cmSynthetic Aperture Sonar 50 x50cmArea:100km2 17000 images(overlapping)Investigating a multiphysics neural network approachBathymetryBackscatterSASPredictionComparing Neural NetworksBathymetryBackscatterSASPredictionPredictionPredictionP