DemodataMjolnir1024
Acquisitionparameters:
Acquisitiontime:
11/416atapproximately15:00localtime
Altitude:
120m
Pixelsizeontheground:
~4cm
DEMusedforgeoreferencing: SRTM30
Boresightcalibration:
Notperformed
Files:
DataLevel
UAVSpeed
Integrationtime/Frameperiod
losby_04_UAV_Mjolnir_SN5001_raw_rad_keystone_smile_spectralbinningx2_bsq_float32.img
Radiance/unrectified
1m/s
39900us/40000us
losby_05_UAV_Mjolnir_SN5001_raw_rad_keystone_smile_spectralbinningx2_bsq_float32.img
Radiance/unrectified
2m/s
19900us/20000us
line5_unrectified.jpg
quickview/unrectified
2m/s
19900us/20000us
losby_05_UAV_Mjolnir_SN5001_raw_rad_keystone_smile_spectralbinningx2_bsq_float32_geo_NN.bsq
Radiance/Rectified
2m/s
19900us/20000us
losby_05_cutout
Radiance/Rectifiedcutout2m/s 19900us/20000us
line5_cutout.jpg
quickview/Rectifiedcutout2m/s 19900us/20000us
line5.jpg
quickview/Rectified
19900us/20000us
ThedatawereacquiredinearlyspringinNorway,overagolfcourt.Duetotheseason,thedeciduous
treeshavenoleafsandtheamountoffreshgreengrassissparse.Thesunislowandtheskywasmostly
cloudy,butduringline05wehadsomedirectsunlightformostoftheline.Inline04,halfthelineiswith
somedirectsunlight.
Sotheacquisitionconditionswerefarfromideal,butthedataqualitycanstillbedemonstratedquite
well.
Wehavemademostlevelsofdataavailableontheftp.
Hereistheprocessingchainperformedonthedatafromrawdatatogeoreferencedatsensorradiance
(ATCOR4processingnotperformed):
Someconsiderationsregardingsignallevel:
Forthe05lineweareflyingat2m/sandwithintegrationtime19.9ms.
Withthisexposuretimeyoudonothaveanysaturationintheimage,andthevegetationpeaksat600DN
(outof4095)intherawimage.ThismeansthatyoucanpotentiallyincreasetheSNRonthevegetation
spectrumbyafactorsquarerootof4095/600=2.6(thiscorrespondsto6.8timeshighersignallevel).
Forthe04lineyoucanget1.84timeshigherSNR.
Toincreasethesignallevel/SNRyoucandothefollowing:
Flywithhighersunanglethanwhatwedidforthisflight
Flyunderclearskyconditions
Flyslowerasthiswillincreasethemaxpossibleexposuretimewiththesamepixelsize
Flyhigher(foragivenspeed),thiswillalsoallowlongerintegrationtime
Youcanalsousehigherthanthetheoreticalvalueaswell,thenyoudogetstretchedpixels,but
youcanselectabiggergridwhenyoudotherectificationlateron.
NOTE:Duringthisflight,wealsouseanequalizationfilterintheentranceaperturewhichreducesthe
amountoflightinthemiddleofthespectralrangetoaround25%(transmissionremainscloseto100%
attheshortestandlongestwavelengths).Byusingthisfilter,youcanintegratelongerwithoutreaching
saturationforthemiddlewavelengthsandtherebygethigherSNRforlongandshortwavelengths.The
filterwasoptimisedforadifferentinstrument,soadedicatedequalizationfilteradaptedtotheMjolnir
1024spectralresponseisunderdevelopment.Thisfilterisanoptionalitem.Forthisflightitwould
actuallyhavebeenbetternottouseitsinceyouarefarfromsaturationonthevegetationwiththefilter
mounted.
ThemainfocuswhendesigningMjolnir1024hasbeentomakepurespectrawithextremelylow
distortions(keystoneandsmile).Toavoidcompromisingthespectralintegrityofthedataduringthe
rectificationandgeoreferencingweselectthefollowingresamplingmethodinPARGE:
Thatmeansthatyouareselectingtheclosestpixelcenterfromyourunrectifiedimagewithrespectto
therectifiedgrid,andnotdoanyresamplingwhichwouldcompromisethespectralquality/integrity.This
givesabitmorecoarse/grainyoutputimage,butitisveryimportantthatyouusethismethodsothat
thespectralintegrityisintact.IfyouusetheBilinearInterpolation,youroutputimagewilllookalot
smoother,butthespectrumwillchange.
Ifyoulookattheunrectifieddatayoucanseesomesmallvibrations,butduetothenearperfect(around
500ns),timetaggingofeachimagewithrespecttotheposition/attitudedata,youdonotseeany
vibrationsintherectifiedimage.Thisisagoodproofofthetimingaccuracy.Theresidualvibrationswill
befurtherreducedbyusingadifferenttypeofpassivedamperswithdifferentstiffness.
Asimpleexampleofaprocessedimagecanbefoundonthenextfewpages.
Sampletarget:Greentrousersongrassbackground.
Detectionimageshownonthenextpages.
AnomalydetectionimageusingstandardCRX.
Zoomimagefromtheredboxisshownonthenextpage.
AnomalydetectionimageusingstandardCRX.
