Transcription
Geocoding Remote Sensing DataRüdiger Gens
Geocoding remote sensing dataOutliney definitionsy cocktail ingredients for geocodingxxxxxspheroiddatummap projectionresamplinginterpolationy various geocoding methods2
DefinitionsGeocoding remote sensing datay geocodingx geometric transformation of an image into acartographic map projectiony georeferencingx relating image coordinates to map coordinates bydefining control points (usually image corners)y geometric correction and image rectificationare sometimes used synonymouslyx geocoding maybe part of geometric correction3
Geocoding remote sensing dataSphere versus spheroidSource: ArcGIS help filey assumption that the earth is a sphere ispossible for small-scale maps (smaller than1:5000000)y to maintain accuracy for larger-scale maps(scales of 1: 1000000 or larger) a spheroid isnecessary4
Sphere versus spheroidGeocoding remote sensing datafbaGeocentriclatitudeGeodeticlatitude5
Geocoding remote sensing dataCommon SpheroidsyyyyyyBessel 1841Clarke 1866, Clarke 1880GEM 6, GEM 10CGRS 1967, GRS 1980International 1924, International 1967WGS 72, WGS 846
Geocoding remote sensing dataDatumSource: ArcGIS help filey defines the position of the spheroid relative to thecenter of the earthy provides a reference frame for measuring locations onthe surface of the earthy defines the origin and orientation of latitude andlongitude lines7
Geocoding remote sensing dataCommon DatumsyyyyyyWorld Geodetic System 1972 (WGS 72)World Geodetic System 1984 (WGS 84)North American Datum 1927 (NAD 27)North American Datum 1983 (NAD 83)European Datum 1950 (ED 50)South American Datum 1969 (SAD 69)8
Map projectionsGeocoding remote sensing datay "orange peel problem"x distortion in the shape, area, distance, or directionof the datay three general typesx conicx cylindricalx planar9
Geocoding remote sensing dataMost common map projectionsUniversalTransverseMercatorAlbers Equal Area ConicPolar StereographicSource: ArcGIS helpLambert Conformal ConicLambert Azimuthal Equal Area10
Geocoding remote sensing dataResamplingy transformation of image coordinates intoprojection coordinates using a mappingfunctionx usually determined as a polynomial fitx accounts for user defined output pixel sizey determination of the resulting pixel value usingan interpolation method11
Standard interpolation methodsGeocoding remote sensing datay Nearest neighbor interpolationx takes pixel value closest to calculated locationx preserves original pixel valuesy Bilinear interpolationx weighted average (2x2 kernel)x smoothing effecty Cubic convolutionx third degree polynomial fit (4x4 kernel)x essentially low-pass filter12
Interpolation using SincsGeocoding remote sensing dataSinc function: sin(x) / x1.210.80.60.40.20-0.2-0.413
Interpolation using SincsGeocoding remote sensing datay theoretically ideal filterx provides error-free interpolation of the band-limitedfunctionsy problem: no function can be at the same timeband-limited and finite-supporty solution: truncationy practical problem: slowest of the slowest as itrequires large kernel sizes14
Weighting functions1.1000001.0000000.900000ScaleGeocoding remote sensing dataWeighting functions0.800000HammingKaiser (alpha 6)0.700000Lanczos0.6000000.5000000.40000015
Weighting functionsGeocoding remote sensing datay Hammingx α usually 0.54y Kaiser α (1 α ) cos(π τx ) for x τHamming( x,τ ,α ) 0else I (α 1 ( x τ ) 2 ) 0Kaiser ( x,τ ,α ) I 0 (α ) 0 for x τelsex where I0(x) is the zeroth order modified Bessel functiony Lanczos sin (π τx ) Lanczos ( x,τ ) π τx 0for x τelse16
Geocoding remote sensing dataCubic B-Splinesy piecewise polynomial function of degree threey very good approximation of sinc functiony generally as fast as cubic convolutionÎ best bang for the buck17
Geocoding by co-registrationGeocoding remote sensing datay image to imagex reference needs to be map projectedy image to mapx map in raster or vector formatx map needs to have map coordinatesy image with measured ground control pointsx ground control points (GCPs) need to be identifiedin the imagex GCPs need to be known in some map coordinatesystem18
Sensor Geometric ModelGeocoding remote sensing datay Sensor Modelx sensor specificx analytical reconstructionof image formation usingorbit and sensorparametersx corrects image globallyx small number of groundcontrol points to improveparametersx DEMZSrGreenwichxw NovYiX19
Examples for sensor modelGeocoding remote sensing datay optical datax Landsat (level 1G)x MODIS (level 1B)x SPOT (level 2A and 2B)y radar datax any beam mode20
Geocoding remote sensing dataGeocoding stepsy relation between image coordinates andgeographic coordinates using image geometryx line / sample Æ latitude / longitudey conversion of geographic coordinates into mapprojected coordinatesx latitude / longitude Æ xmap / ymapx choice of map projection and datum21
Geocoding remote sensing dataGeocoding stepsy determination of a transformation function tomap image coordinates into projectioncoordinatesx usually quadratic, at times cubicx linear least squares polynomial fity resampling using mapping functionx determination of pixel value in the map projectedusing one of the interpolation methods22
Geocoding remote sensing dataExample: Original imageSource: ITC23
Geocoding remote sensing dataExample: Transformed imageSource: ITC24
Geocoding remote sensing dataExample: Geocoded imageSource: ITC25
Light at the end of the tunnel*Geocoding remote sensing datay asf geocodexxxxcurrently under developmentsupports all major map projectionssupports all major datumssupports all standard interpolation methods*for SAR data26
More background informationGeocoding remote sensing datay image processing literaturex medical imagingx astronomyx signal processingy remote sensing data providersx product descriptions for the various satellite imagery27
Geocoding remote sensing dataQuestions28
Geocoding remote sensing data Geocoding steps ydetermination of a transformation function to map image coordinates into projection coordinates xusually quadratic, at times cubic xlinear least squares polynomial fit yresampling using mapping function xdetermination of pixel value in the map projected using one of the interpolation methods