SPOT stereo matching for DTM generation Page 1 SPOT stereo m(2)

This paper presents a matching algorithm for automatic DTM generation from SPOT images that provides dense, accurate and reliable results and attacks the problem of radiometric differences between the images. The proposed algorithm is based on a modified v

A third aspect of our research was the development of algorithms for the automatic detection of blunders.

Figure 1.

Radiometricdifferencesduetoagriculturalactivities(leftpair)andduetocloudsandshadows(right pair).

2. TEST DATA

AstereoSPOTpanchromaticlevel1AmodeloverW.Switzerlandwasacquired.Theinclinationofthesensor’sopticalaxiswas23.4 Rand19.2 Lrespectively,leadingtoaB/Hratioofca.0.8.Theacquisitiondateswere20.7.1988and27.8.1988withsigni cantradiometricdifferencesbetweenthetwoimages,particularlyinagriculturalareas.Figure1showssometypicalim-agepartswithlargeradiometricdifferences.Theelevationrangewas350-3000m.Thefollowingpreprocessingwasappliedtothe original digital images:

reduction of periodic and chess pattern noiseWallis ltering for contrast enhancement

136controlandcheckpointswereusedwithKratky’srigorousSPOTmodel(Kratky,1989b).10ofthepointswereusedascontrolpointswithalinearmodeloftheattituderatesofchange.Thepixelcoordinatesweremeasuredinoneimagemanuallyandtransferredtothesecondonebytemplatematching.TheRMSofthecheckpointswas9-10minplanimetryand6minheight.

3. MODIFIED MPGC

MPGCisdescribedindetailinBaltsavias,1991.Itcombinesleastsquaresmatching(involvinganaf negeometrictransfor-mationandtworadiometriccorrections)andgeometricconstraintsformulatedeitherinimageorobjectspace.Theconstraintsleadtoa1-Dsearchspacealongaline,thustoanincreaseofsuccessrate,accuracyandreliability,andpermitasimultaneousdeterminationofpixelandobjectcoordinates.Anynumberofimages(morethantwo)canbeusedsimultaneously.Themeas-urementpointsareselectedalongedgesthatarenearlyperpendiculartothegeometricconstraintsline.Theapproximationsarederivedbymeansofanimagepyramid.Theachievedaccuracyisinthesubpixelrange.Thealgorithmprovidescriteriaforthedetection of observation errors and blunders, and adaptation of the matching parameters to the image and scene content.InthecaseofmatchingofSPOTimagesthegeometricconstraintswereformulatedasfollows.First,givenameasurementpointinoneoftheimages(templateimage)aheightapproximationisneeded.Iftheexistingapproximationsrefertothepixelcoordinates,thentheheightiscomputedbyusingthepixelcoordinatesinthereferenceimage,thexpixelcoordinateinthesec-ondimageandtheimagetoimagePMFs.ThisheightZisalteredbyaheighterror ingtheheightsZ+ Z,Z– Z,thepixelcoordinatesinthetemplateimageareprojectedbytheimagetoimagePMFsinthesecondimagewheretheyde nethege-ometricconstraintsline.Inthesequel,thisquasiepipolarlinewillbereferredtoasepipolarline.Thecentreofthepatchofthesecondimagewhichisusedformatchingisforcedtomovealongthislinebymeansofaweightedobservationequationoftheform

vc=(x+ x)cosβ+(y+ y)sinβ–p

(1)