Home

Geometric Calibration Plan

1. GCP chips library Figure 4 1 6 NIR and Red Blue SWIR Scene GCP matching The GCP chips extraction and matching is performed independently for each band and camera more exactly for each strip and the result is collected in the scene strip residual file containing the GCPs residuals difference between the observed and predicted computed GCPs coordinates In particular scene strip residual includes for each strip e List of line pixel position for each GCP found by image correlation e List of line pixel position for each GCP predicted by using current version of the ICP e List of flags indicating if the GCPs are used for inversion or not e List of SNR correlation value for each GCP The global level geometric calibration is illustrated in Figure 4 1 7 NIR band and Figure 4 1 8 other bands It is achieved by performing the inversion model over a set of scenes acquired at different times but with the same expected thermal deformations e g with the same temperature and exit time from eclipse within some margins The corresponding GCP residuals of this set of scenes are indicated in Figure 4 1 7 and Figure 4 1 8 as Scene 1 Residuals Scene 2 Residuals Scene N Residuals In line with the baseline approach the sensor Exterior Orientation parameters boresight angles are common among all the bands and these parameters are calibrated using NIR images e g NIR calibratio
2. We propose to use a set of geographic ROIs covering approximately 6 degrees latitude and going from 180 degrees west to 180 degrees east over the 60N to 56S latitude range and all the longitudes in order to skip the world regions without SRTM DEM The LIC files are expected to cover the whole swath across track and to be approximately 800 km long ref N77D7 PV02 US 13 IQC GC CAL CS 108 21 49 Advanced Computer Systems A C S S p A Reference N77D7 PV02 US 13 IQC GC CAL CS 108 ADVANCED COMPUTER SYSTEMS Version 2 2 Date 73 11 2012 The following table and figure show a proposal set of such ROIs and the result of a simulation done by the PF of the coverage of datasets which would be produced in a day on the basis of these ROIs Table 4 2 1 input ROIs The following figure shows the result of the ROI clipping e Red segments segments acquired by the left camera e Green segments segments acquired by the center camera e Blue segments segments acquired by the right camera e Black segments segments resulting from the clipping using the ROIs specified in the Table 4 2 1 itime a mime 3 E a ff Pe ea Ss EEEL aE ie siete aain Ww i 150 100 7a oo PTL 1st Figure Errore Nel documento non esiste testo dello stile specificato 2 1 Simulated coverage of 1 days calibration data ref N77D7 PV02 US 13 IQC GC CAL CS 108 22 49 Advanced Computer Systems A C S S p A Reference N77D7 PV
3. Advanced Computer Systems A C S S p A Reference N77D7 PV02 US 13 IQC GC CAL CS 108 Version 2 2 Date 73 11 2012 Select calibration period lt a Plots USER 5 i cr lt yoxseg 3nd3noQ ynduyT Figure 4 2 8 EstTrend input output data 4 2 3 Calibration development strategy 4 2 3 1 Geometric Calibration validation At the beginning of the commissioning phase the first goal to achieve is the complete validation of the Geometric Calibration facility using real data It includes also the validation of the post launch calibration methods algorithms used within the GC facility It is important to point out that the Geometric Calibration validation activity have to be done by GC developing team since it includes SW debugging and algorithmic implementation aspects Using real data during the commissioning phase the calibration parameter are tuned e g order of deformation polynomial weighing factors search windows etc In this phase the JProcNewScenes processing are properly configured in order to not delete the WDs allowing quality investigation of the generated products ref N77D7 PV02 US 13 IQC GC CAL CS 108 32 49 Advanced Computer Systems A C S S p A Reference N77D7 PV02 US 13 IQC GC CAL CS 108 Version 2 2 Date 73 11 2012 The product reprocessing are used in order process the same input datastrip by using different configuration parameters e Gene
4. N77D7 PV02 US 13 IQC GC CAL CS 108 38 49 Advanced Computer Systems A C S S p A Reference N77D7 PV02 US 13 IQC GC CAL CS 108 CCS Version 2 2 Date 73 11 2012 4 fo ee a gee PP ae gt Oe Fg Sores oP a a ae a mw FA A A ee ee te 0 _ PRAA AA ae EA m Ea r a A es ae OO T Pe fA PE P a a al fog VLEs eek E7 aa ee aa me s 200 t FAF T AT A eee eee a a Pte Af ee AA eee ie fe 47 A nee Ze i eof FF a ee Lf sts ee ee San irigt 7 400 f f j y LS arf P tg te ge ae Pas ees M WY 1 ee att ee aD Wa s b r AAA AA he BOU a ne ARA a Ay A eae roe a ns p a a moe ta ar a oe e per f i ee e aa 1 owe la ee FP tO OT FP ye SE A A a ee ee i g0 a a i eae SOO 1 4 im ye aa E A 1 Ae Se on ad ey Len ea ein a he a i 1 ae A Af Tere AR e 1 Tn y Ta H A an a ni o Per eA e tAE anr e a r E R 1000 2g 400 Bot og 10g 12400 140 Figure 4 2 12 Example of residual plots Statistical evaluation As with any other least square solution statistical methods are used to evaluate the results of a block adjustment We evaluate the precision of the solution by examining the covariances of the parameters as outlined in section 7 2 5 The diagonal elements of the covariance matrix are the parameter variances and the off diagonal elements are the covariances between the parameters variances The relationship between the parameters covariance matrix and th
5. Strips Section 7 2 Obseverd GCP coordinates Predicted GCP coordinates NIR Scene strips Parameter Errors Costrained Inversion Model Section 8 NIR Scene strips Parameter Values Satellite metadata NIR Scene strips Residuals Figure 4 1 3 NIR scene strips calibration performed first ref N77D7 PV02 US 13 IQC GC CAL CS 108 14 49 Advanced Computer Systems A C S S p A Reference N77D7 PV02 US 13 IQC GC CAL CS 108 CIC S ADVAN o a Sey a EA i e AAN RA A TRE Version 2 2 F iM aS maS i oe Keo A j 4 Date 73 11 2012 Figure 4 1 4 illustrates in more detail the Red Blue and SWIR calibration processing flow Again absolute Red Blue and SWIR geo localization errors are computed during the calibration by performing appropriate statistic computation on the GCPs residuals difference between the observed and predicted computed GCPs coordinates Again the GCP residuals are computed on a set of GCPs used to validate the geolocation accuracy performance of the geometric calibration and which are not used in the scene inversion model E Current LoS model Direct georeferencing Section 5 GCP chips library GCP resampling in detector geometry Section 7 1 Level 1C mod etlips Chip matching BI trips gt Scene SWIR strips Section 7 2 Obseverd GC
6. 23 Figure 4 2 11 Mean July Total Cloud Amount The most critical areas are located at equatorial latitudes and at high latitude please note that latitude over 60 degree north are not processed during the winter cloudiness about 80 is expected therefore only 20 of the GCPs can be processed assuming that the probability of an observed Proba V pixel being cloudy is equal to the percentage of cloudiness However it 1s expected that several satellite passes over the same area highly increases the probability to process the GCPs According to JU ROY the probability of there being at least one cloud free ref N77D7 PV02 US 13 lIQC GC CAL CS 108 36 49 Advanced Computer Systems A C S S p A Reference N77D7 PV02 US 13 IQC GC CAL CS 108 petal Loci alot tL AR ae T i ih m Version 2 2 Date 73 11 2012 observation of a PROBA V pixel within a given period is derived from the probability of all of the overpasses being cloudy fe 1 p i where n is the number of satellite overpasses in the period of interest and p is the probability of the i th acquisition being cloudy Considering a period of 3 weeks the number of satellite overpasses is about 10 the PROBA V cycle is 2 days However the cloudiness is correlated over time Here we assume that the cloudiness observations taken at 5 days intervals are slightly uncorrelated the resultant probability of there being at least one cloud free observation as a functi
7. CS 108 19 49 Advanced Computer Systems A C S S p A Reference N77D7 PV02 US 13 IQC GC CAL CS 108 ADVANCED COMPUTER SYSTEMS all SS er a Rd ae TRE Version 2 2 Date 73 11 2012 e T is the temperature ee aia l and are the ideal sensor coordinates e 7 T is the effective focal length o 2 T and yos T are the principal point shift position of the CCD detector central pixel in PP PP the spectral imager SI coordinate frame Ax T and Ay T are the CCD displacements with respect to the ideal sensor I coordinate From PV01 it is clear that there will be multiple temperature readings required spread over the instrument to define the relation to the viewing angles However currently the parameters are expected to depend linearly on a single temperature per camera the temperature or a combination of temperatures having the highest impact due to distortions of the optical elements The following self calibration parameters shall be analyzed and implemented during the project e General polynomial model of degree M M A Ax bande T Da ia T x j l M A Aye T D T x j l Please note that the principal point shift is equivalent to polynomial constant term A preliminary analysis on the simulated data has been done The main outcome is that this general polynomial model gives satisfactory results considering a 8th order of polynomial coefficients However more accurate assessment of t
8. be the one which gives the best fit A pre launch fitting order polynomial tuning is performed by using on ground calibration and measurement data together with the validated thermal elastic model simulations at different temperatures and positions along the orbit and the corresponding results are used as starting fitting order polynomial during the commissioning phase see section 4 2 3 ref N77D7 PV02 US 13 IQC GC CAL CS 108 23 49 Advanced Computer Systems A C S S p A Reference N77D7 PV02 US 13 IQC GC CAL CS 108 Version 2 2 Date 73 11 2012 From PV01 it is clear that there will be multiple temperature readings required spread over the instrument to define the relation to the viewing angles However thermal elastic model simulations will be used in order to define the temperature or a combination of temperatures having the highest impact due to distortions of the optical elements available values O used values required values 10 30 temperature ae deg one ne ee Pee ee O Q e eae 70 Figure 4 2 3 Temperature time exit from exlips matrix in ICP files 4 2 2 Data Management Plan A calibration data management plan which specifies the input and output data flows is part of the calibration plan and it is described in this section The processing chain GenChipsDB is used generate a Geocover Chips GCP Database from the Landsat
9. camera CCD temperature or other currently unknown parameters but large enough to make a reasonable estimate of the calibration parameters The calibration can be divided in 4 main processing chains which can be used to monitor and calibrate the Proba V sensor GenChipsDB is used generate a Geocover Chips GCP Database from the Landsat Geocover data The whole database covering all land between latitude 60N and 56S should be prepared before launch of the satellite JProcNewScenes is used to calculate scene specific calibration parameters boresight angles and focal plane distortion parameters whenever new calibration datasets are available This could be launched automatically with a configurable time span to process any newly available data ref N77D7 PV02 US 13 IQC GC CAL CS 108 8 49 Advanced Computer Systems A C S S p A Reference N77D7 PV02 US 13 IQC GC CAL CS 108 Version 2 2 Date 73 11 2012 EstTrend is used to generate a new ICP file when it is considered necessary This could be set automatically for example once a month or could be run by an operator when it is considered necessary on the basis of the geometric monitoring GeoMonitor is used to monitor the current condition of the PROBA V system from a geometric calibration point of view It can be used to see the calibration situation both in the form of numerical error statistics and graphical plots both scene based and against possible variables such as CCD temper
10. cells increasing respectively from 180W to 180E and 60N to 60S The principal external and internal interfaces are reported in the following Figure 4 2 2 1 which gives a static overview of the GC sub systems input output data ICP es 5 9 a Datastrips L1C scenes dd 32 2 OF NextICP p 5s gv w o a Zso 3m CalReports p gt E Ox lt ec o ge lt Repository 7 ad GCPs chips DEM GCconf es ICPs Ses CalReports cS NextICP O p pT ScenePars Q L SceneResiduals 0 USER Figure 4 2 4 Geometric Calibration Facility main external and internal interfaces The external inputs the current ICP instrument calibration parameter file and the Proba V datastrips to be used for calibration The external outputs are the updated ICP to be used as current ICP until the next time it is calibrated and the CalReport which is the GC part of the calibration verification report ref N77D7 PV02 US 13 IQC GC CAL CS 108 25 49 Advanced Computer Systems A C S S p A Reference N77D7 PV02 US 13 IQC GC CAL CS 108 Version 2 2 Date 73 11 2012 The main internal data are e The GCPs chips and DEM tiles for the georeferencing e GCconfs configuration files e ScenePars sensor Exterior Orientation parameters and Interior Orientation extracted from the geometric calibration of each processed scene e SceneResiduals residuals for each processed scenes used by GeoMonitor The main processing chai
11. ground calibration and measurement data together with the validated thermal elastic model of PV01 For the following calibration campaigns later on the input ICP file is the latest one generated by the GC facility Scene 1 N Parameter Errors Scene 1 N Parameter Values Loop on Level 1C scene Vector of Self calibration parameters to viewing direction converter Section 9 1 Boresight angles Viewing directions Errors End Loop Least square Fitting ICP file Section 9 3 Calibration Report Figure 4 1 5 Temperature variation Calibration and ICP file generation In this diagram Interior Orientation parameters principal point offsets and polynomial CCD distortions are called Self Calibration parameters 4 1 2 2 Optional Geometric Calibration Dataflow The diagrams that follow describe the high level processing flows for optional Geometric Calibration strategy at global level ref N77D7 PV02 US 13 IQC GC CAL CS 108 16 49 Advanced Computer Systems A C S S p A Reference N77D7 PV02 US 13 IQC GC CAL CS 108 Version 2 2 Date 73 11 2012 Current ICP file Viewing direction to Self calibration parameters converter Section 9 2 Level 1C NIR strips Scene Red strips one sirips Scene strips 3 cameras GCP matening Residuals SWIR strips Section 7 SAT metadata Current LoS model
12. 02 US 13 IQC GC CAL CS 108 ADVANCED COMPUTER SYSTEMS ai ey r aa SE Version 2 2 eel ae DRE ie pee ee ae zie Date 73 11 2012 The total size of the L1C data simulated for 1 day is 133 7Gb According to VITO if these were to be compressed with the hdf SZIP functionality their volume should be reduced by approximately 50 These data are processed regardless of their effective contribution to the Geometrical Calibration Note that the GC will automatically give less weight to scenes where estimated precision of estimated parameters is lower and more weight to scenes in under represented conditions Figure 4 2 2 Example of PROBA V one day coverage of the 3 cameras 1 day orbiting with green right camera red center camera blue left camera ICP parameters are stored in the form of a temperature time starting from eclipse matrix so that a full set of distortion parameters can be bi linearly interpolated or extrapolated for every temperature time starting from eclipse pair If temperature and time starting from eclipse are totally correlated all the information in a temperature time starting from eclipse matrix will lie on a single curve If this is almost but not quite the case they will lie on a thin strip Distortions are assumed to depend linearly with temperature and according to a low order polynomial with order lt 3 with time starting from eclipse The order used for the time starting from eclipse dependency will
13. 4 2 3 1 GC CDR RIDs 005 the Geometrical accuracy budget is included in Error Analysis document N77D7 PV02 US 09 IQC GC ARPT ACS 0110 1 1 ref N77D7 PV02 US 13 IQC GC CAL CS 108 2 49 Advanced Computer Systems A C S S p A Reference N77D7 PV02 US 13 IQC GC CAL CS 108 Version 2 2 Date 73 11 2012 27 13 11 2012 Removed sentence rid QR PV02 063 sec 4 2 1 ref N77D7 PV02 US 13 IQC GC CAL CS 108 3 49 Advanced Computer Systems A C S S p A Reference N77D7 PV02 US 13 IQC GC CAL CS 108 Version 2 2 Date 73 11 2012 Table of contents Me MIN RO VIN acces cos cscs vn E R 5 E E E OG E A E EE E E E A E E T ATA 5 Eo APP aD ae ce A A EE E E E E E E T AT 5 A RELERENCES eero E E E 6 ZA Applicable doc MA 10 rereset i areen E EEA OE oem antec EEEIEE EET O AA 6 2 Relrence do une NiS erenn ea E E 6 3 TERMS DEFINITIONS AND ABBREVIATED TERMS ssseesssssssceccsssseeocsssseesosoe 7 4 GEOMETRIC CALIBRATION bisessisceccatessessecenaceascissstessacseusstensnscentscansendeoseuesnsssstocseonts 8 4 1 Geometric Calibration Background cccccccceeseeeseeeesesesseeeseeseseseeseeeeeeseeeeeeeeeeeeeseeeeeees 8 AN My TAS 1G COC E DIS eea A E E ET O doused eteco E toadessenedeunsaeecs 8 41 2 Description of the Method yeaa geeqsdacecaseenusecsaaa nace EEE AEREN ERSE EEEREN EES 9 4 1 2 1 Baseline Geometric Calibration Dataflow ccccceseeessessssssesssssesssesesseesseeseseeeeeeees 12 4 1 2 2 Optional Geo
14. 7D7 PV02 US 5 IQC GC ATBD ACS 0102 2 1 SDD Technical Note Contribution to US Calibration Plan N77D7 PV02 US 14 IQC GC SDD ACS 0103 3 2 ARPT PROBA V IQC GC Technical Note Contribution to US Calibration Plan PROBA TN ACS VITO 0108 1 1 Table 1 Applicable documents 2 2 Reference documents GLOS PROBA V Terms Definitions and Abbreviations N77D7 PROBA V TDA v2_ 11 Junchang Ju and David P Roy The availability of cloud JU ROY free Landsat ETM data over the conterminous United States and globally Remote Sensing of Environment 112 2008 1196 1211 ref N77D7 PV02 US 13 IQC GC CAL CS 108 6 49 Advanced Computer Systems A C S S p A Reference N77D7 PV02 US 13 IQC GC CAL CS 108 ADVANCED COMPUTER SYSTEMS Version 2 2 SS e PL eee 5 Date 73 11 2012 Table 2 Literature references 3 TERMS DEFINITIONS AND ABBREVIATED TERMS General terms definitions and abbreviations used within the scope of this document can be found in GLOS In addition the following are used ACS Advanced Computer Systems A C S S p A ref N77D7 PV02 US 13 IQC GC CAL CS 108 7 49 Advanced Computer Systems A C S S p A Reference N77D7 PV02 US 13 IQC GC CAL CS 108 Version 2 2 Date 73 11 2012 4 GEOMETRIC CALIBRATION 4 1 Geometric Calibration Background 4 1 1 Basic concepts The goal of geometric calibration of an optical sensor system is to model the line of sight for each pixel element of the im
15. C S S p A Reference N77D7 PV02 US 13 IQC GC CAL CS 108 CICS ADVANCED anid shila SYSTEMS La aa ce r a he i 3 i Version 2 2 Date 73 11 2012 Red Blue SWIR Red Blue SWIR Red Blue SWIR Scene 1 strips Scene 2 strips ae Scene N strips Residuals Residuals Residuals Red Blue SWIR Multi Scene Parameter Errors Costrained Inversion NIR Multi Scene Model Parameter Red Blue SWIR Section 8 B Values Multi Scene Parameter Values Current LoS model Red Blue SWIR Red Blue SWIR Red Blue SWIR Scene 1 strips Scene 2 strips Scene N strips Complete Complete aN Complete Residuals Residuals Residuals Figure 4 1 8 Red Blue SWIR global level geometric calibration performed after NIR band calibration The ICP generation method approach is the same of the baseline geometric calibration see correspondingly dataflow reported in the previous section 4 1 2 3 LoS modelling For a given camera C and spectral imager band the viewing direction vector LoS in the detector coordinate system spectral imager SI coordinate frame see PF DPM is parameterized in the following way fband _ ge T Ax band C T Ay x band C ye C T yond C T yom T y T T _ pm C T E pines C r where e i 0 N 1 where N number of pixels of spectral imager band and camera C ref N77D7 PV02 US 13 IQC GC CAL
16. Geocover data The whole database covering all land between latitude 60N and 56S should be prepared before launch of the satellite and therefore the data flow relative to the GenChipsDB it is not described here in the calibration plan ref N77D7 PV02 US 13 IQC GC CAL CS 108 24 49 Advanced Computer Systems A C S S p A Reference N77D7 PV02 US 13 IQC GC CAL CS 108 Version 2 2 Date 73 11 2012 The GCP chips dataset is stored in file system based repository It is organized in directories by Landsat WRS2 paths and rows For example the image chips corresponding to Landsat path 188 and row 42 are in directory 188 042 Each directory contains see SDD for more details e Chips directory contains the actual image chips in 91x91 float raw binary format e GCPlist txt ASCII file in table format giving the list of chip filenames and chip center geodetic and map coordinates e GCPscene txt ASCII file in tag value format giving various info including geocover scene metadata and chip extraction parameters The 3 arc second pixel spacing corresponds to 90m SRTM V3 Digital Elevation Model DEM is used It covers 80 of land from the latitude 60 North to 56 South see SDD for more details All the files are in a common srtm_v3 SRTM Data GeoTiff directory and are called Z COL ROW TIF where 1 lt COL lt 72 indicates the column and 1 lt ROW lt 24 the row number in a grid of non overlapping 5 x 5
17. P coordinates Predicted GCP coordinates Red Blue SWIR Scene strips Parameter Errors Costrained Inversion Model Red Blue SWIR Section 8 Scene strips Parameter Values Boresight angles NIR Scene strips Parameter Satellite metadata Values Red Blue SWIR Scene strips Residuals Figure 4 1 4 Red Blue SWIR scene calibration performed after NIR band calibration From PV01 it is clear that there will be multiple temperature readings required spread over the instrument to define the relation to the viewing angles However currently the parameters are expected to depend linearly on a single temperature per camera the temperature or a combination of temperatures having the highest impact due to distortions of the optical elements so the parameters can be linearly fitted against camera temperature to determine the most probable linear law according to Figure 4 1 5 A weighted least squares fit is used according to what specified in ATBD Section 9 ref N77D7 PV02 US 13 lIQC GC CAL CS 108 15 49 Advanced Computer Systems A C S S p A Reference N77D7 PV02 US 13 IQC GC CAL CS 108 Version 2 2 Date 73 11 2012 It is important to point out that the input current LoS model in the first calibration campaign shall be generated starting from the starting ICP file It is assumed that this starting ICP file is externally generated outside the GC facility using on
18. Reference N77D7 PV02 US 13 IQC GC CAL CS 108 Version 2 2 Date 73 11 2012 PROBA V IQC GC Technical Note Contribution to US Calibration Plan 13 11 2012 Prepared by ACS GC team 12 10 2011 Verified by A Arledler 12 10 2011 ref N77D7 PV02 US 13 IQC GC CAL CS 108 1 49 Advanced Computer Systems A C S S p A Reference N77D7 PV02 US 13 IQC GC CAL CS 108 ADVANCE D COM PUT ER SYSTE hil 3 Version 2 2 Date 73 11 2012 Distribution List ww a s oo C a kenes vo Sd o P Paper copy E Electronic version Document change record 10 01 2011 Version submitted to the PDR 28 01 2011 Version proposed to address the Sections 4 x PDR Rids on the document 58 03 2011 PDR 045 PDR 062 and PDR Sections 4 x 097 2 0 14 06 2011 Version submitted to the CDR GC DDR RIDs 012 Sec 4 2 3 and 4 2 4 013 Sec 1 1 014 Sec 4 1 2 2 015 Sec 4 1 2 017 Sec 4 1 2 1 021 Sec 4 2 1 022 file naming convention Sec 4 2 2 024 Sec 4 2 2 025 independent camera calibration 026 Sec 4 2 3 2 and Sec 4 2 3 3 027 Sec 4 2 3 2 PV02 CDR RIDs 062 Sec 4 1 2 3 064 Sec 4 2 1 Completed all the TBW Sections in response to all the related RIDs 1 12 10 2011 Version proposed to address the GC CDR RIDs 020 Sec 4 2 3 1 050 CDR Rids on the document Sec 4 2 1 053 Sec 4 2 2 Fig 4 2 4 054 Sec 4 2 3 1 056 Sec 4 2 3 2 059 Sec 4 4 072 Sec 4 2 3 2 074 Sec 4 2 3 2 092 Sec 1 1 113 Sec
19. aging system This is usually performed pre flight in laboratory conditions where precise measurements enable to characterize the various aspects of the system Due to possible launch effects distinct operational conditions and thermal deformation a post launch self calibration is performed on a regular basis e g every 60 days to ensure geometric system stability PROBA V is a small satellite which provides only passive thermal control onboard so temperature can play a major role in the geometric calibration process This means that in order to ensure the total geometric system accuracy as specified in the PROBA V requirements a quite complex geometrical model is needed The in flight geometrical calibration shall be able to estimate and monitor on a regular basis the boresight misalignment angles and focal plane deformations for the 3 cameras of the PROBA V sensor and there dependence on temperature and orbit position These geometrical calibration parameters shall be used to update the ICP Instrument Calibration Parameter file which is used by the Processing Facility in order to guarantee the geometrical system accuracy of the system corrected products One of the principal design features of the GC subsystem is the division of the Proba V LIC calibration datasets in scenes of a configurable length of approximately 800 km along track for which the calibration parameters can be assumed to be approximately constant with respect to
20. alue of a given data type representing a specific type of information e g a radiometric value in the form of a DN or a reflectance value etc lon lat coordinates etc The lines of observations returned by a strip are transmitted to earth and stored in successive rows of a band Each image pixel in each row and hence each column in the band corresponds to a given detector pixel ref N77D7 PV02 US 13 IQC GC CAL CS 108 9 49 Advanced Computer Systems A C S S p A Reference N77D7 PV02 US 13 IQC GC CAL CS 108 ADVANCED COMPUTER SYSTEMS Bee ree a o A ae eR Version 2 2 Date 73 11 2012 Resolution 100 350 m SWIR DETECTOR 3 butted strips of 1024 pixels each SWIR 570 165Dnm 200 700 m Band Number nm Resolution 1 30m 2 525 605 30m 3 30m ETM Spectral bands 4 30 900 30m 550 1750 Jm 6 1040 1250 60m 7 2090 2350 30m 520 900 15m Figure 4 1 1 Rationale of the PROBA V Landsat Geocover bands selection for the geometric calibration Details are given in Section GCP Chips Generation The calibration is made by registering each band directly and independently with corresponding Landsat Geocover band chips e NIR band is used to calibrate the boresight angles and NIR band focal plane distortions e Other bands are used to calibrate the respective focal plane distortions The rationale of the PROBA V Landsat Geocover bands selection for the geometric calibration is reported in Figure 4 1 1 more detail
21. an be increased by changing the threshold value for the Moravec Interest Operator see ATBD 6 5 4 without significantly affecting the quality of the GCPs e g eliminating low contrast features like water bodies desert dry lands etc in the GeoCover scenes Of course only a fraction of them are effectively used for the calibration seasonal effects clouds etc but it 1s expected to have a sufficient number of GCP used for the calibration Please note that 1 scene selection can be done for the parameter calibration e g selecting only the scenes cloud free or few clouds 2 MMIO extracts chips well distributed over the image avoiding that most of the chips are concentrated over a small region of the image ref N77D7 PV02 US 13 lIQC GC CAL CS 108 11 49 Advanced Computer Systems A C S S p A Reference N77D7 PV02 US 13 IQC GC CAL CS 108 Version 2 2 Date 73 11 2012 3 calibration parameters errors are computed via error propagation ATBD section 8 2 5 and these errors shall be considered during the least square fitting it means that in the case that a scene have less GCPs in same part of the field of view larger errors are expected and the contribution to the final fitting is less important In addition during the calibration campaign several scenes at the same temperature latitude shall be processed therefore the final calibration parameters shall be a weighted average value with a reduced error with respect
22. ature and latitude both with current ICP values and with optimal ones 4 1 2 Description of the method This section gives an overview of the method used for the geometrical calibration It contains material mainly extracted from the ATBD However this duplication is needed in order to make the Geometric Calibration Plan document self consistent The PROBA V instrument is made up of three camera s Each camera has two sensors the VNIR sensor is associated with the visible bands and the SWIR sensor with the infra red band A detector is associated with a full swath line of observations in a given band While the VNIR sensor contains three detectors that each corresponds to a single strip the SWIR sensor contains only one and it is made up of three strips Summarizing the PROBA V instrument has the following characteristics e 3 camera s 1 e the Left camera the Center camera and the Right camera e 4 bands and hence 4 detectors camera i e BLUE RED NIR and SWIR e 2 sensors camera 1 e the VNIR sensor associated with the BLUE RED and NIR bands and the SWIR sensor associated with the SWIR band e 6 strips camera 1 e BLUE RED NIR SWIR1 SWIR2 and SWIR3 An image consists of one or more image bands each consisting of a given number of columns and rows of image pixels e BLUE RED and NIR strips are composed by 5200 pixels e SWIR1 SWIR2 and SWIR3 strips are composed by 1024 pixels Each pixel corresponds to a v
23. bration parameters and their errors to verify that e the geometric distortions are as expected by the current model e the software is working correctly In addition as detailed in 4 2 3 during the commissioning phase specific analysis is envisaged in order to perform calibration parameter tuning e g order of deformation polynomial etc It is important to point out that a pre launch model selection and calibration parameter tuning is performed by using on ground calibration and measurement data together with the validated thermal elastic model of PVO1 and the corresponding results are used as starting model parameters setting during the commissioning phase ACS and VITO shall contribute to the pre flight calibration by supporting the definition of the minimal set of parameters to retrieve from the on ground calibration During the commissioning phase it is verified if the baseline calibration strategy as described in section 4 1 2 1 1s fully adequate for PROBA V calibration or the global solution optional calibration strategy as described in section 4 1 2 2 is necessary Initially during the commissioning phase geometric calibration monitoring with the geo monitor application and ICP updates will probably be quite frequent As the time goes on the percentage of data used for calibration and the frequency of calibration monitoring and ICP updates can be reduced according to trends and error statistics noted during monitoring
24. chosen so that the orientation parameters can be considered constant within the scene and that the scene is large enough to make a reasonable estimate of the parameters The parameters can therefore be estimated for each scene separately and the parameter trend can be plotted vs temperature and or orbit position If necessary the global solution optional calibration strategy is possible by a relative simple customization of the SW the Sensor Exterior and Interior Orientation parameters are estimated considering a set of multiple scenes acquired at the same temperature and time starting from eclipse The advantage of the global solution is that the number of ground control points increases allowing an accurate estimation of the parameters in case of failure of the baseline calibration strategy However it is expected that the baseline calibration strategy is fully adequate for PROBA V calibration In fact sufficient number of GCPs about 1600 are expected in a scene 800 km long acquired over land for each camera This estimation is done by using the updated version of the MMIO SW originally developed for Landsat TM ETM for Proba V as described in section 6 We used 8 Landsat scenes over land obtaining about 100 GCP chips for each scene and per band Considering that a Proba V scene of 800 km x 800 km is composed by about 16 Landsat scenes the number of GCP chips per band is about 1600 Please note that if needed the number of GCPs c
25. commissioning phase The clean up policy is described in SDD Section 4 13 1 The ProcNewScenes processing chain is composed by 2 tasks SceneGep and ScenelInv e SceneGep finds the positions of the geocover chips in the ProbaV images by image cross correlation in the focal plane near the predicted position e ScenelInv finds optimal calibration parameters according to a weighted least squares criterium including iterative outlier removal The complete static description of the ProcNewScenes input output interfaces is given the Figure 4 2 6 ref N77D7 PV02 US 13 IQC GC CAL CS 108 28 49 Advanced Computer Systems A C S S p A Reference N77D7 PV02 US 13 IQC GC CAL CS 108 Version 2 2 Date 73 11 2012 SceneResiduals e ScenePars er Current ICP Datastrip L1C scenes Logging Exit Code Figure 4 2 6 ProcNewScenes input output data GeoMonitor is used to monitor the current condition of the PROBA V system from a geometric calibration point of view It can be used to see the calibration situation both in the form of numerical error statistics and graphical plots both scene based and against possible variables such as CCD temperature and latitude both with current ICP values and with optimal ones It is a commandline based interactive tool that can be launched whenever it is considered necessary to monitor the geometric calibration situation in the form of graphica
26. commissioning phase are given Table 4 3 1 Nominal case V stands for the ICP validation Month 1 Month 2 Month 3 Table 4 3 2 Degraded case 1 V stands for the ICP validation Month 1 Month 2 Month 3 ref N77D7 PV02 US 13 IQC GC CAL CS 108 43 49 Advanced Computer Systems A C S S p A Reference N77D7 PV02 US 13 IQC GC CAL CS 108 Version 2 2 Date 3 2012 Table 4 3 3 Degraded case 2 V stands for the ICP validation Month 1 Month 2 Month 3 First ICP generation Data collection ref N77D7 PV02 US 13 IQC GC CAL CS 108 44 49 Advanced Computer Systems A C S S p A Reference N77D7 PV02 US 13 IQC GC CAL CS 108 ADVANCED COMPUTER SYSTEMS Version 2 2 Date 3 1 2012 4 4 Operational scenario A preliminary vicarious calibration plan for the operational scenario is given Table 4 3 1 nominal case V stands for the ICP validation the ICP monitoring is performed periodically at the end of second week of each month in the middle of the ICP generation period Month 1 Month 2 Month 3 Month 4 Month 5 Month 6 ref N77D7 PV02 US 13 IQC GC CAL CS 108 45 49 Advanced Computer Systems A C S S p A Reference N77D7 PV02 US 13 IQC GC CAL CS 108 ADVANCED COMPUTER SYSTEMS Version 2 2 Date 3 1 2012 Table 4 3 2 degraded case V stands for the ICP validation during the Month 3 the ICP monitoring forces for a generation of a new ICP file Month 1 Month 2 Month 3 Month 4 Mo
27. composed by NIR scene calibration and Red Blue SWIR scene calibration ref N77D 7 PV02 US 13 lQC GC CAL CS 108 Advanced Computer Systems A C S S p A Reference N77D7 PV02 US 13 IQC GC CAL CS 108 ADVANCED COMPUTER SYSTEMS O SSS eh fo if 3 AA oee a M gt s RN St ae SRE Version 2 2 Date 73 11 2012 Figure 4 1 3 illustrates in more detail the NIR scene calibration processing flow Absolute NIR geo localization errors are computed during the NIR calibration by performing appropriate statistic computation on the GCPs residuals difference between the observed and predicted computed GCPs coordinates The GCP residuals are computed on a set of GCPs used to validate the geolocation accuracy performance of the geometric calibration and which are not used in the scene inversion model In particular scene residual includes for each strip e List of line pixel position for each GCP found by image correlation e List of line pixel position for each GCP predicted by using current version of the ICP e List of line pixel position for each GCP predicted by using the best fit inversion model on that scene e List of flags indicating if the GCPs are used for inversion or not e List of SNR correlation value for each GCP GCP chips Current library LoS model GCP resampling Direct georeferencing N R in detector geometry Section 5 Section 7 1 Level 1C NIR Chip matching Scene
28. cover datasets from different WRS2 paths at higher latitudes see ATBD Section 8 3 3 for more details ICP parameters are stored in the form of a temperature time starting from eclipse matrix so that a bi dimensional fitting 1s performed The polynomial order used for the temperature and for the time starting from eclipse dependency will be the one which gives the best fit A pre launch fitting order polynomial tuning is performed by using on ground calibration and measurement data together with the validated thermal elastic model simulations at different temperatures and positions along the orbit for a representative subset of CCD elements e g in the form of the PV0O1 System Performance Simulator csv files and the corresponding results are used as starting fitting order polynomial during the commissioning phase Search Window ref N77D7 PV02 US 13 IQC GC CAL CS 108 33 49 Advanced Computer Systems A C S S p A Reference N77D7 PV02 US 13 IQC GC CAL CS 108 Version 2 2 Date 73 11 2012 The starting ICP accuracy defines the size of the search window and therefore the initial performance enlarging the search window implies an increment of the processing time However the main inaccuracy is related to the boresight misalignment angles determined at the satellite launch and therefore not predictable during the pre flight calibration During the Geometric Calibration validation the initial boresight angle is roughly est
29. d ICP file operation with GeoMonitor see Section 4 2 4 In any case if we have reached this point the weekly routine operations have finished Too rapid inter orbit variation Is the disagreement between current and next ICP due to an inter orbit deformation which is more rapid that expected By viewing trend plots with GeoMonitor one can see if the long term trend in deformations is different from expected If for example linear trends have been modeled and the higher order terms are significant over the current period this should be visible 95 confidence intervals can be used to help determine if the difference is due to random effects or not More detailed analysis If the problem is not due to too rapid inter orbit variation a more detailed analysis 1s necessary which could be done by local personnel or ACS depending on the problems encountered Enough data The differences between the deformations estimated from data in the current period and deformations from current ICP may be because not enough data in the current period has been processed This may be for example because in the current period there have been too many clouds at a particular latitude In this case the 95 confidence limits visible with GeoMonitor on the final fit for that latitude should be wider than normal Request more data If we want to generate a new ICP file but do not have enough data in the current period to generate it with the requested accuracy i
30. e on ground measurement accuracy is performed by the Sensitivity Analysis Figure 4 2 2 shows the principle components of the Sensitivity Analysis tool 3 The values of parameters for the LoS model necessary for the mapping process have to be defined position of the satellite actual look direction sensor parameters and errors ref N77D7 PV02 US 13 lIQC GC CAL CS 108 39 49 Advanced Computer Systems A C S S p A Reference N77D7 PV02 US 13 IQC GC CAL CS 108 Version 2 2 Date 73 11 2012 2 Subsets of parameters of the LoS model can be selected for the error simulation The others will be assumed to be exact Q The variation of the selected parameters is performed via Gaussian noise given by the standard deviation o and or covariance matrix of the parameters Monte Carlo method 4 Additional inputs for the LoS model are the sensor description and the processing definitions For example to process not only one look direction but the whole FOV range 5 The outputs of the geo location error are given in terms of Root Mean Square Error RMSE and the Standard Deviation STDV Provided a parameter range is processed for example the whole FOV the error can be displayed in error curves RMSE STDV Graphic Sensor Type Processing Defs ERNENSANDE lie e SPECIFICATION Figure 4 2 13 Flow chart of sensitivity analysis tool the green arrows show a possible
31. ges should still be available on local disk More detailed analysis If the problem is not due to bad data a more detailed analysis is necessary which could be done by local personnel or ACS depending on the problems encountered GeoMonitor can be used to see where errors are greatest or GCPs are found on specific scenes and strips Enough data If the data is enough even though some data is bad the daily routine check can in any case be considered successfully finished In case of doubt anticipate the weekly routine check ref N77D7 PV02 US 13 IQC GC CAL CS 108 48 49 Advanced Computer Systems A C S S p A Reference N77D7 PV02 US 13 IQC Version 2 2 Date 3 2012 4 4 2 Weekly operations Next current ICPs agree Do the distortions predictions based on data from the current month agree with those of the current ICP based on data from the previous month After running EstTrend to fit and calculate statistics on data from the current month it can be compared to the current ICP Current and next ICPs can be considered in agreement if their distortion values are within an excepted margin and or their estimated 95 confidence regions overlap Note that GC maintains statistics and confidence margins of each calculated ICP file Time to generate new ICP Is it time to generate a new ICP Nominally a new operational ICP is generated each month So if all is well and the time has come we can make the already generate
32. he bundle adjustment is run a qualitative evaluation is performed as a first step in which the operator examines graphical representations of the adjustment s output in order to understand broad trends and to catch obviously bad inputs e g GCP chips outliers not correctly detected Finally statistical analysis techniques are used to protect against bad observations and to quantify the quality of the adjustment Qualitative evaluation The purpose of qualitative evaluation is to allow the operator to understand the properties of the adjustment and to diagnose any problem It is difficult to recognize trends or outliers by looking at pages of numbers instead since this is geometric problem it is best addressed by graphical displays of the solution outputs The plots proposed for qualitative evaluation of results of the bundle adjustment are of the image residuals as in Figure 4 2 1 The image residuals should point in random directions and have comparable sizes Residuals all pointing in the same direction either parallel or radially with respect to the center of the image indicate the presence of a systematic effect These effects may indicate uncorrected errors or an image parameter that has been weighted too highly and not allowed to adjust A residual that is larger than its neighbors or points in the opposite direction to those near it indicates a bad measurement outliers nor correctly removed during the bundle adjustment phase ref
33. he polynomial model can be done by using on ground calibration and measurement data together with the validated thermal elastic model of PVO1 pre launch and by using real data during the commissioning phase The model selection strategy is described in Section 4 3 2 The following table gives an overview of the set of parameters to be calibrated for a given PROBA V scene approximately at constant temperature Exterior Orientation Parameters Boresigh angles for each camera 3 angles x 3 cameras 9 Interior Orientation Parameters Principal Point for each camera and strip 2 PP x 3 cameras x 6 strip 36 Self Calibration Parameters Polynomial model for each camera and strip of M x 3 cameras x 6 strip 18 M degree M ref N77D7 PV02 US 13 IQC GC CAL CS 108 20 49 Advanced Computer Systems A C S S p A Reference N77D7 PV02 US 13 IQC GC CAL CS 108 Version 2 2 Date 73 11 2012 The number of calibration parameters to extract from a single strip scene via GCPs TPs correlation is 6 M where M is the degree of the self calibration polynomial 4 2 Geometric Calibration Plan 4 2 1 Overview Once in orbit any data collected in nominal mode and processed to form L1C products covering land in the 60N to 56S latitude range can be used for Geometric Calibration Initially during the commissioning phase all available data should be used to determine as soon as possible first estimates of onboard cali
34. imated in debug mode by analysing the data by GC developing team and manually setting a suitable large search window Once the first estimation of the boresight angle is done the search window size is reduced tuned according to the real data In the case that the GC accuracy requirements are not satisfied by the baseline Geometric Calibration approach nominal scenario alternative approaches shall be analyzed and implemented degraded scenario 1 Firstly increasing the number of the GCPs per scene and camera extracted by the MMIO tool by changing the threshold values for the Moravec Interest Operator ATBD see 6 5 4 2 Secondly if the first solution is still not adequate by using the global solution via a relative simple customization of the SW see Section 4 1 2 2 The first approach is straightforward to implement and requires only 3 4 days in order to generate the new set of GCPs chips A pre launch definition of the new threshold values for the Moravec Interest Operator is done in order to speed up the GCPs chips generation The second approach requires a reorganization of the SW tasks in new processing chains as described in SDD which needs about 1 month for developing and SW testing 4 2 3 2 First ICP generation After the consolidation of the calibration parameters tuning the fist ICP file shall be generated by using the tuned calibration parameters set In order to generate the first set of ICP file it is ex
35. itable set of test data to verify evaluate results for inter camera calibration which is not available The number of ground control points is crucial for the end result of the per camera approach it is expected that the number of GCPs per scene and camera extracted by the MMIO tool is adequate However if necessary the number of ground control points can be increased e by changing the threshold value for the Moravec Interest Operator ATBD see 6 5 4 e by using the global solution is possible by a relative simple customization of the SW see above 4 1 2 1 Baseline Geometric Calibration Dataflow The diagrams that follow describe the high level processing flows for baseline Geometric Calibration processing algorithms In these figures within each box indicating the principal algorithm processing step is reported the corresponding Section where the algorithmic method 1s described In all of these diagrams the following naming convention 1s used e LoS model set of parameters used to model the sensor Exterior Orientation parameters boresigh angles and Interior Orientation parameters principal point offsets polynomial CCD distortions for ref N77D7 PV02 US 13 IQC GC CAL CS 108 12 49 Advanced Computer Systems A C S S p A c C S Reference N77D7 PV02 US 13 IQC GC CAL CS 108 aaa zi COM all ER svete MS Version 2 2 Date 73 11 2012 each spectral band and camera The LoS model is the result of an overall geome
36. l plots and text The static description of the GeoMonitor input output interfaces is given the Figure 4 2 7 ref N77D7 PV02 US 13 IQC GC CAL CS 108 29 49 Advanced Computer Systems A C S S p A Reference N77D7 PV02 US 13 IQC GC CAL CS 108 ADVANCED COMPUTER SYSTEMS a SSS he OL It AA ear ee A e AAR eat TRE Version 2 2 OMS Oe AL Date 73 11 2012 SceneResiduals A1ojyisoday Figure 4 2 7 GeoMonitor input output data ScenePar contains for each scene the complete ICD is reported in SDD e Scene parameter values sensor Exterior Orientation parameters and Interior Orientation parameters extracted from the geometric calibration of a single scene e Scene parameter errors errors associated to the Scene parameter values covariance matrix e Scene condition parameters parameters averaged or representative for the whole scene with respect to which the orientation parameters may vary such as TimeOutofEclipse temperatures from various onboard sensors latitude date and time of scene etc SceneResidual includes for each scene the complete ICD is reported in SDD List of line pixel position for each GCP found by image correlation List of line pixel position for each GCP predicted using the current ICP file List of flag indicating if the GCPs are used for inversion or not List of SNR correlation value for each GCP All ScenePars and SceneResiduals generated from a single LIC file a
37. metric Calibration Dataflow ccccccseesesssssssssseessesessssssesseseseeeeeseeeeees 16 BP Zc Lon mod NAIA ea r eE E 19 42 Geomerie Calibration Plan sesser E O 21 AA A ON eare E E EE E E E E E E T E E 21 422 Data Mandpement Plan secite sirrien anan a E AE EEEE OE EE EOE 24 4 2 3 Calibration development strategy ccccccccccsesesssseessseeeeeeeeeeeseseeseseseeeeeeeeeeeeeeeeeeeeegs 32 A231 Geometric Calibration Vali atl ON gags tsas25secaatesvecanteisaea weesnsasneioanastessnsasieioseew EE 32 Ae Ue IO P pener i esei aE E REE E E 34 Be Oo New Ge PGi ANION a5cccc sac scseaaacgnaedaeesaessosed ce orian EE E E 38 42A CAMO Fat Old TOM OLIN a casaeesaiseasepacrsteenaeieavanheaiavonzeseasetieeh vexadstinticdansballsadsiabauaesbatadadsiatedsnstatc 38 AD Commissione PAN tesa AEE AEE ERNA EEEE EA EEAO REEE 43 E Oe E E E E E 6 E A En eee AA A T 45 EEA NM Daly operons eona E E ened carte ane eed 48 4A2 Weekly Operations arssoiisreiiieeieeriit oriei pa carne Ae ENESE AEE EOE SEAE Ei eia 49 4 49 ref N77D 7 PV02 US 13 lQC GC CAL CS 108 Advanced Computer Systems A C S S p A Reference N77D7 PV02 US 13 IQC GC CAL CS 108 Version 2 2 Date 73 11 2012 1 INTRODUCTION 1 1 Scope This document gives the geometric calibration plan for the PROBA V GC subsystem This document gives a summary of the Geometrical Calibration method and elaborates the geometric calibration activities after the launch of the satellite in term
38. n is performed first The output of the global level geometric calibration are e Multi scene Parameter values sensor Exterior Orientation parameters boresigh angles and Interior Orientation parameters principal point offsets polynomial CCD distortions extracted from the geometric calibration of the set of input scenes ref N77D7 PV02 US 13 IQC GC CAL CS 108 17 49 Advanced Computer Systems A C S S p A CCS tla COM PUTER mii Ee Hi E Date Reference N77D7 PV02 US 13 IQC GC CAL CS 108 Version 2 2 13 11 2012 e Multi scene Parameter errors errors associated to the Multi scene Parameter values Error analysis e Scene strip complete residuals a separate file for each scene the scene strip residuals see above including the list of line pixel position for each GCP predicted by using the global level best fit inversion model NIR NIR NIR Scene 1 strips Scene 2 strips Scene N strips Residuals Residuals Residuals NIR Multi Scene Parameter Errors Costrained Inversion C urrent Mon LoS model NIR Section 8 Multi Scene Parameter Values NIR NIR NIR Scene 1 strips Scene 2 strips Scene N strips Complete Complete Complete Residuals Residuals Residuals Figure 4 1 7 NIR global level geometric calibration performed first ref N77D7 PV02 US 13 IQC GC CAL CS 108 18 49 Advanced Computer Systems A
39. n systematically used during the geometric calibration is JProcNewScenes The JProcNewScenes is a Java program which uses the JODI framework in order to launch and monitoring the ProcNewScenes processing chain ProcNewScenes is used to calculate scene specific calibration parameters boresight angles and focal plane distortion parameters whenever new calibration scenes are available This is launched automatically with a configurable time span to process any newly available scenes The JProcNewScenes input output dataflow is fully described by using the activity diagram reported in Figure 4 2 5 ref N77D7 PV02 US 13 IQC GC CAL CS 108 26 49 Advanced Computer Systems A C S S p A Reference N77D7 PV02 US 13 IQC GC CAL CS 108 Version 2 2 Date 73 11 2012 Systematic processing Recursively scheduled Datastrip user reprocessing Backup repository WorkingDirectory clean up Recursively scheduled process failed proc ess completed Figure 4 2 5 JProcnewScenes activity diagram Some details about the activity diagram are given below e the JProcNewScenes is launched automatically with a configurable time span by using a Crontab demon e the first task of the JProcNewScenes is to check the availability of a datastrip scene in a configured Input Basket directory in the case that no datastrip is available the JProcNewScenes terminates e if a datatrip is available JProcNewScenes creates a Working Director
40. nteed that all the necessary data can be acquired during a single day In addition the user can check that there is an appropriate distribution of GCP across FOV by using GeoMonitor tool see Section 4 2 4 e visual check by using the GCP residual plot e g Figure 4 2 12 e looking at the polynomial coefficients confidence limit e g Figure 4 2 16 In fact it is expected large errors for some polynomial coefficients in the case there are not GCPs in a part of the FOV Figure 4 2 9 show the mean annual cloudiness from the ISCCP dataset http isccp giss nasa gov composed by observations in the period ranging from July 1983 through June 2006 ISCCP D2 198387 200806 Hean Annual J 180 120 Bo 1270 180 Total Cloud Amount 23 Figure 4 2 9 Mean annual Total Cloud Amount In order to illustrate the cloudiness variability within the different seasons the monthly cloudiness in January and July are reported in Figure 4 2 10 and 4 2 11 respectively ref N77D7 PV02 US 13 IQC GC CAL CS 108 35 49 Advanced Computer Systems A C S S p A Reference N77D7 PV02 US 13 IQC GC CAL CS 108 ADVANCED COMPUTER SYSTEMS ik oie an a AA E R ii Version 2 2 Prad A r F T tiran i Date 73 11 2012 ISCCP 02 198367 200806 Hean January At At Total Cloud Amount CE Figure 4 2 10 Mean January Total Cloud Amount ISCCP D2 198307 200806 Hean July 90 180 70 60 g 120 Total Cloud Amount
41. nth 5 Month 6 ref N77D7 PV02 US 13 IQC GC CAL CS 108 46 49 Advanced Computer Systems A C S S p A Reference N77D7 PV02 US 13 IQC GC CAL CS 108 CIC S ADVANCED COMPUTER SYSTEMS ea ee Date 3 1 2012 More SceneGCP Bad data detailed Too rap More OK analysis inter orbit detailed analysis Anticipate weekly generate check ew ICP Generate Request new ICP more data Figure 4 3 1 Daily and weekly routine monitoring cycles ref N77D7 PV02 US 13 IQC GC CAL CS 108 47149 Advanced Computer Systems A C S S p A Reference N77D7 PV02 US 13 IQC Version 2 2 Date 73 11 2012 The routine geometric monitoring will be organized in very quick and basic operations to do every day and slightly longer operations to do every week If the daily operations indicate that something is wrong the weekly operations may be anticipated If either the daily or weekly operations indicate that something is wrong but it is not clear what is wrong more detailed analysis will be necessary Both the routine and more detailed analysis will be well documented in the software user manual once the software has been implemented and the most practical way to do the checks has been determined via simulations In any case the preceding diagrams give a 1 overview of the routine operations and the text paragraphs below give a 1 description of the steps If during commissioning or operations certain pr
42. oblems with the system appear specific checks for these problems will probably be intensified On the contrary if other problems never occur the checks for other problems will become less frequent or more superficial 4 4 1 Daily operations SceneGCP OK Have the L1Cs received in input been processed as expected and have the expected amount of GCPs been found This can be verified both from errors and warnings in the log files and by viewing the scene summary statistics with GeoMonitor see Section 4 2 4 The number of GCPs found for each scene gives a good indication SceneInv OK Has the scene inversion occurred as expected Again errors and warnings in the log files can give a 1 indication Number of outliers RMSE of residuals and estimated RMSE of estimated distortion again visible in scene summary statistics with GeoMonitor good indications see Section 4 2 4 If all is as expected in the above 2 steps the daily checks have finished If not Bad Data Are the encountered problems simply due to bad data Extensive cloud cover is the most probable cause but portions of data may be unusable for other reasons If the percentage of scenes with few or no GCPs or high RMSE is low or as expected no extra checks are necessary If cloud or bad data statistics or quicklooks are easily available from the PF they may be consulted to see if the degraded functioning of GC is due to these problems Otherwise the full resolution ima
43. on of the cloudiness is given in Table 4 2 2 Cloudiness Pwe Equivalent Cloudiness _ 9 ___On a ooo o o o o O e 0 02 o o o Table 4 2 1 Mean probability of there being at least one cloud free observation as a function of the cloudiness in the case of n 3 the number of independent satellite overpasses The last column contain the Pone expressed in terms of equivalent cloudiness The first ICP file is generated by using EstTrend which performs the requested fit and it is able to show the plot of the result with the associated confidence error figure which is the first element for the ICP validation In order to perform a more complete ICP validation as requested the GeoMonitor tool is used The main input of the GeoMonitor is the SceneResidual which includes for each scene the list of line pixel position for each GCP predicted by using current version of the ICP in this case the pre launch ICP file The GCP residuals are also computed on a set of GCPs which are not used in the scene inversion model ground check points The first ICP file can be fully validated according to the following steps 1 Select a set of scenes not used for the generation of the ICP file to be validated 2 Use the GeoMonitor with the generated set of SceneResiduals files Once the ICP is fully validated the file is copied in the Output basket for the distribution ref N77D7 PV02 US 13 IQC GC CAL CS 108 37 49 Advanced Compu
44. pected to process a set of data at different latitudes and temperatures e g about 10 different latitude temperature spread over the latitude range from 60N to 56S In addition for each couple latitude temperature several scenes per camera about 800 km long according to ATBD are requested in order to reduce the calibration parameter errors please note that the errors goes like 1 sqrt n where n is the number of scenes A preliminary analysis is reported in the PROBA V IQC GC Technical Note Contribution to US Calibration Plan ARPT An accurate analysis estimation of n can be done only during the commissioning phase using EstTrend tool which provides the confidence intervals of the temperature time starting from eclipse least square fitting algorithmic details are given in the ATBD section 9 From a simulation done by PF according to the input ROI reported in Table 4 2 1 it is expected that approximately 75 scenes per camera per day have to be processed ref N77D7 PV02 US 13 IQC GC CAL CS 108 34 49 Advanced Computer Systems A C S S p A Reference N77D7 PV02 US 13 IQC GC CAL CS 108 Version 2 2 Date 73 11 2012 However it 1s important to point out that an useful scene shall contain a sufficient number of GCPs therefore it shall be almost cloud free or partially covered by clouds therefore it is expected that only part of the acquired scenes can be used for the geometric calibration This means that it is not guara
45. ral polynomial of degree N e Weighting factors 1 0 where o is the expected standard deviation of the calibration parameters e Search windows e Other TBC The generated results with different parameters are compared and analyzed by using GeoMonitor tool as described in Section 4 2 4 Deformation Polynomial A preliminary analysis on the simulated data has been done in the case of general polynomial model The main outcome is that this general polynomial model gives satisfactory results considering a 8th order of polynomial coefficients This reference has been used in defining the model A more accurate pre launch order polynomial tuning is performed by using the starting ICP file which is externally generated outside the GC facility from the on ground calibration and measurement data together with the validated thermal elastic model of PV01 The corresponding results are used as starting point during the commissioning phase Weighting factors A pre launch definition of the a priori weight for each calibration parameters and for GCP measurements are used as starting values during the commissioning phase e The determination of the a priori weight for each calibration parameters shall be done by using on ground calibration and measurement data together with the validated thermal elastic model of PVO1 e An analysis of the a priori weight for GCP measurements is done by using 375 image pairs extracted from overlap of geo
46. re grouped together in a single directory and generically called a SceneVal The naming convention for the SceneVal directory is see SDD ref N77D7 PV02 US 13 IQC GC CAL CS 108 30 49 Advanced Computer Systems A C S S p A Reference N77D7 PV02 US 13 IQC GC CAL CS 108 Version 2 2 Date 73 11 2012 lt mission gt SCENEVAL lt camera gt lt datetime gt lt duration gt _ lt purpose gt lt version gt Where lt mission gt gt PV1 lt camera gt gt L C or R to indicate left center or right camera 1 byte lt datetime gt gt YYYYMMDD_HHMMSS datetime of segment start as found in LIC name 15 bytes lt duration gt gt duration of segment in seconds 3 bytes lt purpose gt gt O T C to indicate operational test commissioning phase 1 byte lt version gt gt integer unique file version identifier 3 bytes The GeoMonitor allows the user to choose what to see scene trend or sensitivity case scene choose scene show scene residuals before fit show scene residuals after fit case trend show trend case sensitivity show sensitivity analysis EstTrend is used to generate a new ICP file when it is considered necessary EstTrend is run by an operator when it is considered necessary on the basis of the geometric monitoring see Section 4 2 3 The static description of the EstTrend input output interfaces is given the Figure 4 2 8 ref N77D7 PV02 US 13 IQC GC CAL CS 108 31 49
47. s can be found in ATBD GCP chips generation section In the previous version of the ATBD document an alternative calibration approach was proposed first calculating camera boresight angles and NIR band focal plane distortions using a database of georeferenced image chips generated from band 4 of the Landsat Geocover dataset and then calculating focal plane distortions of other bands from band to band registration However the inter band calibration approach has been discarded mainly for the difficulties of matching NIR to other four spectral bands of the surface areas with variable spectral characteristics as pointed out in RID PDR 066 ref N77D7 PV02 US 13 IQC GC CAL CS 108 10 49 Advanced Computer Systems A C S S p A Reference N77D7 PV02 US 13 IQC GC CAL CS 108 Version 2 2 Date 73 11 2012 Sensor Exterior and Interior Orientation parameters are expected to depend on camera CCD or TMA temperature and on the time starting from eclipse Two different strategies for estimating the Sensor Exterior and Interior Orientation parameters are foreseen see MOM PM6 e Baseline calibration strategy at scene level e Optional calibration strategy if needed at global level This option is not currently implemented but it is possible by a relative simple reconfiguration of the production chains see SDD Assumptions and Limitations The basic assumption of the baseline calibration strategy is that a scene size can be
48. s of e calibration data management plan which specifies the input and output data flows e calibration strategy for the GC model validation and ICP generation and validation e calibration activity planning during and after commissioning phase The commissioning plan elaborates the activity planning for the nominal case and for the two main degraded cases Please note that the Geometric Calibration Plan document contains material extracted from the ATBD and SDD However this duplication is needed in order to make the document as much as possible self consistent 1 2 Applicability It is an ACS contribution to section 5 of the US Calibration plan US 13 US CAL GC This document has to be provided at PDR in draft version at CDR in almost definitive version and should be updated to account for any changes at successive milestones ref N77D7 PV02 US 13 IQC GC CAL CS 108 5 49 Advanced Computer Systems A C S S p A Reference N77D7 PV02 US 13 IQC GC CAL CS 108 Version 2 2 Date 73 11 2012 2 REFERENCES All applicable and reference documents are listed in the SOW N77D7 PV02 PM 19 IQC GC SoW vl 0 draft 2 1 Applicable documents The documents that have been used for the preparation of underlying document are SOW N77D7 PV02 PM 19 IQC GC SoW Proba V PV02 User Segment Geometrical Calibration Statement of Work CPL N77D7 PV02 US 13 US CAL PROBA V US Calibration Plan ATBD Algorithm Theoretical Baseline N7
49. selected configuration the red ones possible selections A fundamental requirement for the bundle adjustment is that it be accurate Accuracy cannot be determined by examining the solution since a solution can be very consistent within itself but still not accurate The accuracy can be determined by computing statistical average and covariance of the GCPs residuals In addition to the analysis at scene level multiple scenes monitoring is very useful including e An indication of which scenes are within specification with current ICP file ref N77D7 PV02 US 13 lIQC GC CAL CS 108 40 49 Advanced Computer Systems A C S S p A Reference N77D7 PV02 US 13 IQC GC CAL CS 108 CCS ADVAN a D anid PUT ER tL Mi 5 5 i E i Version 2 2 Date 73 11 2012 e Plots of various calibration parameters and their 95 confidence errors against various external parameters including temperature exit time from eclipse date time e Plots of expected absolute localization accuracy against various external parameters including temperature exit time from eclipse date time Plot of GCP residuals considering only the set of GCPs which are not used in the scene inversion model ground check points e Plots of estimated absolute geolocation errors within a processed scene with current ICP and with optimal calibration parameters scene residuals with current ICP parameters and optimized scene parameters The most effective plots to show
50. t may be sufficient to shift the start date of the estimation period back a little if the long term variations are not too rapid If this is not the case one can try requesting more data by interleaving the current ROIs with other rows of ROIs ref N77D7 PV02 US 13 IQC GC CAL CS 108 49 49 Advanced Computer Systems A C S S p A
51. ter Systems A C S S p A Reference N77D7 PV02 US 13 IQC GC CAL CS 108 Version 2 2 Date 73 11 2012 4 2 3 3 New ICP generation During the operational scenario new ICP files are generated according to a specified frequency in order to take into account seasonal thermal deformation variations By default a monthly calibration campaign is performed however a careful monitoring of the validity of the current ICP file shall be done over the time and in case of the accuracy requirements violations a new ICP shall be generated The ICP monitoring is performed by monitoring over the time the scene residuals of the new acquired scenes and in particular monitoring the line pixel position for each GCP predicted by using current version of the ICP with respect to the line pixel position found by image correlation The GCP residuals are also computed on a set of GCPs which are not used in the scene inversion model ground check points The validation of the new ICP file is exactly the same procedure used for the validation of the first ICP file described in the section above by using the GeoMonitor with the generated set of SceneResiduals files 4 2 4 Calibration monitoring After running a bundle adjustment a careful evaluation of its results will be performed in order to be sure that all the specifications and requirements are meet and that the results are valid that is not contaminated by bad measurements or assumptions After t
52. to the parameters extracted from a single scene Please note that the PF interpolates the CG parameters stored in the ICP file according to temperature time exit from eclipse therefore no discontinuities are foreseen The main expected advantage of this method is that the computationally expensive part of the calibration processing can be done individually for each scene without making large assumptions on which known environment parameters the distortions depend If the distortions actually depend on position in the orbit or camera TMA temperature or have a particular drift in time this dependency can be noted from plots of scene calibration parameters the sensor model can be modified and new calibration parameters can be computed without repeating the computationally expensive part of the processing This method would not be the most effective if the distortions where to depend only on CCD temperature exactly as expected and the latter were to oscillate rapidly so that the scene length would have to be too short to work But this scenario does not seem probable at the moment It is worth noting that the current calibration approach foresees that each camera is calibrated separately whereas in the previous version of the ATBD document all the three cameras are calibrated together considering Tie Points in the inter camera overlap regions The multi camera calibration approach has been discarded see MOM PM06 mainly because it requires a su
53. tric calibration campaign performed over a set of PROBA V scenes more specifically the LoS model 1s computed from the current ICP file and thus includes temperature time starting from eclipse depended fitting e Scene parameter values sensor Exterior Orientation parameters boresigh angles and Interior Orientation parameters principal point offsets polynomial CCD distortions extracted from the geometric calibration of a single scene e Scene parameter errors errors associated to the Scene parameter values Error analysis e Calibration Report contains the errors relative to the boresight angles and to the CCD pixel viewing directions The sensor Exterior Orientation parameters boresight angles are common among all the bands and these parameters are calibrated using NIR images Current ICP file Viewing direction to Self calibration parameters converter Section 9 2 Current LoS model NIR Scene Calibration first step NIR Scene Parameter SAT metadata Errors Parameter GCP chips Values library loresight angles for each camera Red Blue SWIR Scene Parameter Values Red Blue SWIR Red strips Level 1C eee a Red Blue SWIR Scene p Scene 3 cameras Parameter SAT metadata Errors Red Blue SWIR NIR Scene Scene strips strips Residuals Residuals Figure 4 1 2 Scene Calibration
54. will be chosen when they have been implemented and tried with realistic data The following are some mock examples and will be included in the GeoMonitor tool simulated linear in temp quadratic in latitude with variable gaussian noise 6 0 input residuals 6 6 estimate theoretical E d B 2 5 8 5 6 simulated parameter value 5 4 dee 5 0 4 0 5 0 6 1 7 1 5 0 9 1 1 1 temperature deg 0 ag4r136 5 15688 Figure 4 2 14 example plot of a calibration parameter with 95 confidence against temperature ref N77D7 PV02 US 13 IQC GC CAL CS 108 41 49 Advanced Computer Systems A C S S p A Reference N77D7 PV02 US 13 IQC GC CAL CS 108 acs atiii D ComM oll ER svete MS Teng Version 2 2 Date 73 11 2012 estimated NIF distortions in focal plane measured FSeconf residuals a SICy est FSecont SICx est FSseconf expected distortions pix 0 O 1 1 2 0 4 4 0 5 0 6 O 0 8 0 4 1 SICy striplenth 0 0936105 56 7455 Figure 4 2 15 example plot of NIR distortion in focal plane Figure 4 2 16 Example of Strip0 SICy output errors and 95 confidence limits ref N77D7 PV02 US 13 IQC GC CAL CS 108 42 49 Advanced Computer Systems A C S S p A Reference N77D7 PV02 US 13 IQC GC CAL CS 108 ADVANCED COMPUTER SYSTEMS Version 2 2 Date 3 1 2012 4 3 Commissioning plan A preliminary vicarious calibration plans for the
55. y WD and moves the input datatrip from the Input Basket to the WD in addition the current ICP file is copied in the WD from Repository e if an error occur during the processing chain ProcNewScenes the WD is not removed for investigation e after the completion of the processing chain the generated products ScenePars and SceneResiduals are exported in the repository directory the input datastrips are moved to the backup repository and then the WD is removed optionally ref N77D7 PV02 US 13 IQC GC CAL CS 108 27149 Advanced Computer Systems A C S S p A Reference N77D7 PV02 US 13 IQC GC CAL CS 108 Version 2 2 Date 73 11 2012 The processed scenes are stored in the backup repository for a configured time period according to a pre defined cleaning up policy That gives the user the possibility to perform the data reprocessing by simply moving the selected scenes from the backup repository to the input basket The data reprocessing is an important feature which is expected to be extensively used during the commissioning phase for the GC model verification testing and parameter tuning In addition also Working Directories not removed after the completion of the processing chain are kept within the system for a configured time period according to a pre defined cleaning up policy This feature is very important especially for the Geometrical Calibration testing investigation and model verification using real data during the

Geometric Calibration Plan

Contents

Download Pdf Manuals

Related Search

Related Contents