Updated Results of Calibration and Validation of PRISM and AVNIR-2

Transcription

1 ALOS PI Symposium 2008, Rhodes Island, Greece 3-7 November, 2008 Updated Results of Calibration and Validation of PRISM and AVNIR-2 Takeo Tadono*, Masanobu Shimada*, Hiroshi Murakami*, Junichi Takaku**, Sachi Kawamoto**, and Akira Mukaida** * Earth Observation Research Center (EORC), Japan Aerospace Exploration Agency (JAXA) tadono.takeo@jaxa.jp ** Remote Sensing Technology Center of Japan (RESTEC) Pan-sharpened image of Rhodes, Greece generated by PRISM and AVNIR-2

2 Outline Introduction Data acquisition status Release / update note of processing software at EOC, JAXA Geometric Calibration PRISM: Pointing alignment parameter Time trend of geometric correction accuracy Relative CCD alignment AVNIR-2: Time trend of geometric correction accuracy Band-to-band registration Radiometric Calibration and Image Quality PRISM: Stripe noises reduction JPEG block noises reduction High radiometric gain operation results AVNIR-2: Field Of View (FOV) radiometric calibration Conclusions

3 ALOS Daichi Advanced Land Observing Satellite (ALOS) (Advanced Land Observing Satellite) Jan. 24, 2006: Launch by H-IIA #8 from TNSC Nov. 3, 2008: 2.8 year (1,012 days) after launch Star Tracker PALSAR 8.9m Data Relay Antenna (DRC) [Data rate: 240Mbps] GPS Antenna Mission objectives: - Cartography (1:25,000 scale), - Regional environment observation, - Disaster monitoring, and - Resources surveying. 2.9m PRISM AVNIR-2 Solar Array Paddle 22m Velocity Nadir - ALOS is in good health condition: Fuels, battery and power generation - July and August 2008: Inclination control operations were successfully done. PRISM : Panchromatic Remote-sensing Instrument for Stereo Mapping AVNIR-2: Advanced Visible and Near Infrared Radiometer type 2 PALSAR: Phased Array type L-band Synthetic Aperture Radar

4 ALOS Basic Observation Scenario PRISM (Descending) 9 One global coverage annually (OB1 triplet; OB2 selected areas) 9 2 cycles (2 x 46 days) required for each region (+/-1.2deg. pointing angle) 9 Timing based on cloud statistics, seasonality and sun elevation AVNIR-2 (Descending) 9 One global coverage annually (0deg. pointing) 9 One observation within 2 cycles 9 Timing based on cloud statistics, seasonality and sun elevation PRISM (green: OB1, yellow: OB2) PALSAR (Ascending / Descending) 9 Asc.: 2-3 global coverage annually (Summer FBD34deg.; Winter FBS34) Global InSAR coverage every 2 yrs Pol-InSAR campaigns every 2 yrs 9 Desc.: One global ScanSAR coverage annually Intensive ScanSAR sites * Observation Scenario can be find on web AVNIR-2 PALSAR Desc. (ScanSAR) PALSAR Asc. (FBD34) Basic Observation Scenario (Cycle19: Apr Jun 8, 2008) * Basic Observation Scenario is discussing and authorizing in ALOS Mission Operation Planning Board, which consists of power users (METI, GSI, MoE, MAFF, JCG) and data distributor. It will be held on each six months. EORC Earth Observation Research Center

5 Acquisition Status of PRISM and AVNIR-2 PRISM (Nadir view) AVNIR-2 Acquired area Cloud coverage < 2% Spatial coverage: 51% OB1 OB2 Spatial coverage: 64% * OB1 (triplet)+ob2 (N70km+B35km) Image coverage map of PRISM and AVNIR-2 based on the basic observation scenario (From May 16, 2006 to Jul. 24, 2008 (Rev. 20))

6 Update of PRISM Processing Software at EOC Major updated information of PRISM processing software for standard products at EOC, JAXA See Date (CP) PRISM Pointing Alignment Parameters (AP) updated (Version 16 of Sep 24, 2008) (CP) PRISM Pointing AP updated (Version 15 of July 24, 2008) (CP) PRISM Pointing AP updated (Version 14 of May 26, 2008) (CS) Filter processing program added: Reduction of JPEG block noises (CP) PRISM Pointing AP parameters updated (Version 13 of March 17, 2008) (CP) PRISM Pointing AP updated (Version 12 of January 16, 2008) (CP) PRISM Pointing AP updated (Version 11 of October 29, 2007) (CS) Update radiometric correction algorithm of Level 1 processing: Reduction of stripe noises (CP) PRISM Pointing AP updated (Version 10 of Aug 29, 2007) (CP) PRISM Pointing AP updated (Version 9 of July 17, 2007) (CP) Parameters of geometric model changed: Relative CCD alignments (Version 3) (CP) Parameters of geometric model changed: Relative CCD alignments (Version 2a) (CP) PRISM Pointing AP updated (Version 8 of April 20, 2007) (CP) PRISM Pointing AP updated (Version 7 of March 22, 2007) (CS) Filter processing program added: Reduction of stripe noises (CP) Parameters of geometric model changed: Relative CCD alignments (Ver. 2), Sensor alignment (CP) PRISM Pointing AP updated (Version 6 of February 20, 2007) (CP) PRISM Pointing AP updated (Version 5 of January 31, 2007) (CP) PRISM Pointing AP updated (Version 4 of Oct.8, 2006) Operational phase start CP: Change Parameter, CS: Change Specification

7 Update of AVNIR-2 Processing Software at EOC Major updated information of AVNIR-2 processing software for standard products at EOC, JAXA See Date (CP) AVNIR-2 alignment angles (θy) updated (the geometric correction table) Temporal changes of sensor alignment (and STT coordination) were reflected (10-30m) (CP) Table of the radiometric correction updated Field Of View (FOV) calibration: FOV noise, gain-mode differences and small-scale noise (+/-2%) (CP) Parameters of geometric model changed Changed parameters: Coordination conversion rotation angle, and alignment (θx,θy,θz) (CP) AVNIR-2 alignment angles (θx,θy,θz) updated (0,0,0) -> evaluated values prior to launch Operational phase start CP: Change Parameter

8 Geo Cal PRISM Alignment Parameter (AP) ALOS AOCS system and coordination PRISM CCD coordinate for Nadir For Forward view For Backward view Pointing det. parameter Pointing alignment parameter STT STT STT attitude reference frame Attitude determination parameter ECI(J2000) coordinate STT coordinate - Precise Orbit Determination from TAC - Precise Attitude Determination (PAD) - Pointing Alignment Parameter (AP) Evaluation of variation during recurrent, seasonal change, and temporal change. > It is better to use high latitude and night time GCPs EOC Pointing AP is basically updating every two months i.e. time gap is occurred between observed date and processed date. Two months time gap! Ex) #16 AP Release: Sep. 25, 2008 Evaluation: Mar. 18 Sep. 17, 2008 > Go / No go Valid period: July 17 Sep observation PPDS - Standard product - Evaluation -GCP - Evaluation - Sensor alignment estimation EORC -GCP - Evaluation - Sensor parameters More detail will be presented by Mr. Takaku on tomorrow = We recommend the data order will be better to submit two months after observation date.

9 Geo Cal PRISM Geometric Correction Accuracy Averaged error in X (m) Averaged error in Y (m) /01/07 06/01/07 08/01/07 10/01/07 12/01/07 01/31/08 04/01/08 06/01/08 08/01/08 10/01/08 04/01/07 06/01/07 08/01/07 10/01/07 12/01/07 01/31/08 04/01/08 06/01/08 08/01/08 10/01/ Observation date Observation date Averaged geometric errors of nadir looking radiometer of PRISM L1B2 (left: X (pixel) direction, and right: Y (line) direction). STDV of error in X (m) STDV of error in Y (m) /01/07 06/01/07 08/01/07 10/01/07 12/01/07 01/31/08 04/01/08 06/01/08 08/01/08 10/01/08 04/01/07 06/01/07 08/01/07 10/01/07 12/01/07 01/31/08 04/01/08 06/01/08 08/01/08 10/01/08 Observation date Observation date Standard deviations of geometric errors of nadir looking radiometer of PRISM L1B2 (left: X, and right: Y). Time trends of geometric correction accuracy of PRISM/N L1B2 (as of Oct. 31, 2008) Averaged error: Absolute geometric correction (system correction) accuracy Period: April 22, 2007 September 3, 2008; 370 scenes, 4,210 GCPs as check points

10 Geo Cal PRISM Relative CCD Alignment FWD 142 scenes 2,910 GCPs X_σ = 0.66 pixel Y_σ = 0.69 pixel NDR 202 scenes 3,906 GCPs X_σ = 0.63 pixel Y_σ = 0.62 pixel X Residual [ pixel] X Residual [ pixel] Serial CCD pixel address Serial CCD pixel address Y Residual [pixel] Y Residual [pixel] Serial CCD pixel address Serial CCD pixel address BWD 194 scenes 3,657 GCPs X_σ = 0.64 pixel Y_σ = 0.69 pixel X Residual [ pixel] Serial CCD pixel address Y Residual [pixel] Serial CCD pixel address GCP residuals back-projected on L1B1 images vs. pixel address of CCD units Stability of relative geometric accuracy of PRISM (latest version 3 released on June 2007): On-orbit self calibration: CCD alignment parameters accuracy on image space No significant degradations of accuracy from latest version Interior orientation parameter: keeping sub-pixel level within 0.7 pixel (STDEV)

11 Geo Cal AVNIR-2 Geometric Correction Accuracy Averaged geometric errors of AVNIR-2 L1B2 compared between before and after Oct. 22, (Left: only 0deg. pointing, and right: any pointing from to -41.5deg.) Sensor alignment (Y) of AVNIR-2 looks like change with observation date since launch AVNIR-2 alignment (Y, line direction) is changing with observation date. It may be including attitude coordination variation when the products were processed by Onsite Precision. PRISM s alignment is adjusting by the pointing AP. Sensor alignment parameters were released on Oct. 22, 2008 that are depending on the observation date (four months period). Correction values are +2.2 to -2.4 pixels. The alignment (X, pixel direction) does not change. The errors are caused by quantization of pointing angle setting.

12 Geo Cal AVNIR-2 Band-to-Band Registration Nearest Neighbor (NN) - Potential ambiguity is within 1 pixel Ave(STDEV) X Y Band Band Band Cubic Convolution (CC) Ave(STDEV) X Y Band Band Band Averaged error in X (pixel) Averaged error in X (pixel) Pointing angle (deg.) Poninting angle (deg.) NN B1 x NN B2 x NN B4 x CC B1 x CC B2 x CC B4 x Averaged error in Y (pixel) Pointing angle (deg.) Stability of band-to-band registration (latest version parameter released on Dec. 2006): 2007: 28 scenes, various pointing angles (-41.5 to +41.5deg.), and resampling methods Band 3 is as base image Automatic image matching technique (least square matching) Averaged error in Y (pixel) Pointing angle (deg.) NN B1 y NN B2 y NN B4 y CC B1 y CC B2 y CC B4 y

13 Results of PRISM/AVNIR-2 Geo Cal Standard Product PRISM 1B2 AVNIR-2 1B2 Results as of Sep. 29, 2007 (Public*) Geometry Absolute Accuracy (RMS): using 1,390 GCPs Pixel (X) Line (Y) Distance Nadir 6.5m 7.3m 9.8m Forward 8.0m 14.7m 16.7m Backward 7.4m 16.6m 18.1m Relative Accuracy (1σ) 3 radiometers 1.9m 2.3m 3.0m Geometry (-41.5 to deg. pointing) Pixel (X) Line (Y) Distance Absolute Accuracy (RMS) 106m 19m 108m Relative Accuracy (1σ) 4m 4m 6m Results as of Nov. 3, 2008 (cont evaluated) Geometry (Jun. 22, 2007-Oct. 2, 2008) Absolute Accuracy (RMS) Pixel (X) Line (Y) Distance Nadir 6.2m 5.2m 6.1m using 2,712 GCPs, 268 scenes Forward 3.2m 6.6m 5.7m using 1,271 GCPs, 136 scenes Backward 6.0m 7.0m 7.2m using 2,533 GCPs, 252 scenes Relative Accuracy (1σ) 3 radiometers 1.5m 1.9m 2.4m Geometry (-41.5 to deg. pointing, all period) using 1,035 GCPs, 54 scenes Pixel (X) Line (Y) Distance Absolute Accuracy (RMS) 60.9m 96.6m 73.4m Relative Accuracy (1σ) 3.4m 7.7m 8.4m Only 0deg. pointing Absolute Accuracy (RMS) 71.1m 7.5m 69.6m * Latest ALOS calibration result can be find at in English in Japanese

14 Radio Cal PRISM Stripe Noise Reduction Odd-Even pixel and inter-ccd unit difference were large sometimes in PRISM images We assume PRISM sensor itself is stable and the error is caused by insufficient frequency of the darkcurrent downlink (optical black i.e. offset error) We estimate the dark current statistically using each scenes 1. Inter-CCD unit difference (offset) is corrected by overlap samples (32 pixels) after the default radiometric correction The correction coefficients are tuned to keep mean radiance of all CCD unit 2. Odd-Even pixel difference (offset) is corrected by statistics of the Even minus neighboring two Odd samples in each CCD 3. Above statistics are processed in each one of five line-blocks, and correction offsets are linearly interpolated by the line number Irregular and high-contrast samples are excluded in the statistics Block 1 Block 2 Block 3 Released on October 19, 2007 Block 4 Block 5 CCD 1 CCD overlap A) Odd N-1 PRISM image B) Even N CCD 2 CCD 3 C) Odd N+1 Dif = B (A + C)/2 CCD 4 Statistics in each segment Evaluation results will be presented by Mr. Mukaida in Poster Session on Thursday

15 Radio Cal PRISM Stripe Noise Reduction ALPSM Before After

16 Radio Cal PRISM Stripe Noise Reduction ALPSM Forward Before After

17 Image Quality PRISM JPEG Block Noise Released on April 1, 2008 Before block noise reduction Sometime, block noises were appeared especially in homogeneous areas. After block noise reduction Block noises reduced still keeping image textures. Post processing filter to reduce block noises has been installed to operational software PRISM data is compressing by JPEG into satellite to reduce data downlink rate (960Mbps > 240Mbps) Post filtered processing algorithm was developed to reduce ONLY block noises by JPEG It have been installed in JAXA processing software for standard products (Level 1B1 and 1B2) on Apr 1, 2008 Some parameters tuning may be necessary Reference: Izumi Kamiya and Genya Saito, Reduction of JPEG and other noise for ALOS PRISM image, Proc. 28th Asian Conference on Remote Sensing (ACRS), Nov , Evaluation results will be presented by Mr. Mukaida in Poster Session on Thursday

18 Image Quality PRISM High-Gain Operation Due to requests from image interpretation fields, PRISM have been operated with highradiometric gain setting (Gain 1) as test operation in Japan and Alaska 8bit is not satisfied for image interpretation; PRISM has 4 gain setting (Gain 1 to 4) Default gain: Gain 2 or 3 over land area Gain 1 test operations were conducted in winter season over Japan; compared with previous one Gain 2 (Left image, Observation date: Dec 24, 2007) Saturation rate: 0.01% (5,899/58,519,725pixel) IMG-05-ALPSMW O1B1 W Gain 1 (Right image, Observation date: Feb 8, 2008) Saturation rate: 0.23% EORC (134,032/58,522,452pixel) Earth Observation Research Center IMG-05-ALPSMW O1B1 W

19 Image Quality PRISM High-Gain Operation Gain 2 (Observation date: Dec 24, 2007) IMG-03-ALPSMW O1B1 W Gain 1 (Observation date: Feb 8, 2008) IMG-03-ALPSMW O1B1 W Histogram on gain 1 shows more broad distribution > Better contrast for gain 1 image Higher saturation rate for gain 1 image as pre-expectation > It causes blooming and smear noises

20 Image Quality PRISM High-Gain Operation Due to requests from image interpretation fields, PRISM have been operated with highradiometric gain setting (Gain 1) as test operation in Japan and Alaska 8bit is not satisfied for image interpretation; PRISM has 4 gain setting (Gain 1 to 4) Default gain: Gain 2 or 3 over land area In the case of Alaska, there are not enough images of same RSP, pointing angle, and seasons without cloud Gain 2 (Jul. 25, 2008) Gain 1 (Sep. 9, 2008) Covered by hazes? Percent Percent DN Gain 2 (Jul. 25, 2008) Saturation rate: 0.0% (0/14,116,863pixel) IMG-04-ALPSMB O1B1 B DN Gain 1 (Sep. 9, 2008) Saturation rate: 2.84% (399,534/14,064,960pixel) IMG-04-ALPSMB O1B1 B clouds

21 Image Quality PRISM High-Gain Operation Example of Alaska Gain 1 observation (Sep. 11, 2008) Percent 9 6 Percent DN Gain 2 (Apr. 24, 2007) Saturation rate: 0.0% (0/30,349,296pixel) IMG-05-ALPSMF O1A F DN Gain 1 (Sep. 11, 2008) Saturation rate: 0.0% (0/30,333,648pixel) IMG-05-ALPSMF O1A F

22 Radio Cal AVNIR-2 FOV Calibration: Concept RGB Image 2007/ 11/13 Radiance (or DN) Unrealistic inter-channel difference Released on May 9, 2008 L1B radiance = Real data + FOV noise + small-scale noise (e.g., o/e, cal table error..) Real data + FOV anomaly Real TOA radiance Pixel (FOV; ~70km) Pixel (FOV; ~70km) 1. FOV noise Corrected by cross-calibration with MODIS (using a directional function of MODIS TOA reflectance) Temporal change is described using the internal lamp of AVNIR-2 2. Gain-mode difference Gain-modes 2 and 3 are corrected using the lamp data 3. Small-scale noise (<~0.5DN) Corrected by small scale average of smooth & bright area (polar snow fields) Line average plot of Band 3 on 2007/11/13 L1A L1B Corr. tables Pixels

23 Radio Cal AVNIR-2 FOV Calibration Antarctic 2007/11/19 Band 3,2,1 RGB image old new Band-1~4 line plot Differences between old and new are +/- 2%

24 Radio Cal AVNIR-2 FOV Calibration Libyan Des. 2007/07/04 Band 3,2,1 RGB image old new Band-1~4 line plot

25 Conclusions I introduced overall calibration results of PRISM and AVNIR-2 as a part of operational calibration, in particular, 1) Time trends of geometric correction accuracies of PRISM and AVNIR-2, - PRISM: pointing alignment parameter updates (each two months) * Alignment variation model is developing in EORC for high level products - AVNIR-2: sensor alignment parameter updates (each four months?) 2) Confirmed relative geometric accuracies of PRISM and AVNIR-2, 3) Implemented reductions of stripe noises and JPEG block noises of PRISM, 4) Gain 1 test operation of PRISM: winter season in Japan will be operated, and 5) Radiometric table updated for AVNIR-2: FOV calibration. We are continuously carrying out evaluations of data qualities, and will be updated parameters / algorithm to keep / improve the accuracies as operational calibrations. For more information of ALOS, JAXA/EORC : New images, data acquisition plan and technical documents For data search and order, ALOS User Interface Gateway (AUIG) All archived data can be searched with Guest account Bird s-eye view of Rhodes Island, Greece generated by PRISM DSM with pan-sharpened image.

26 Outline Introduction Data Acquisition Status Basic Observation Scenario Acquisition status in the world Geometric Calibration AVNIR-2: Band-to-band registration Sensor alignment evaluation Time trend of geometric correction accuracy PRISM: Relative CCD alignments Sensor alignment variation model Time trend of geometric correction accuracy Validation of Generated DSMs by PRISM Summary

27 Geo Cal PRISM Geometric Correction Accuracy Averaged error in X (m) Averaged error in Y (m) /01/07 06/01/07 08/01/07 10/01/07 12/01/07 01/31/08 04/01/08 06/01/08 08/01/08 04/01/07 06/01/07 08/01/07 10/01/07 12/01/07 01/31/08 04/01/08 06/01/08 08/01/08 Observation date STDV of error in Y (m) /01/07 06/01/07 08/01/07 10/01/07 Observation date Averaged geometric errors of nadir looking radiometer of PRISM L1B2 (left: X (pixel) direction, and right: Y (line) direction). STDV of error in X (m) /01/07 06/01/07 08/01/07 10/01/07 12/01/07 01/31/08 Observation date 04/01/08 06/01/08 08/01/ /01/07 01/31/08 Observation date Standard deviations of geometric errors of nadir looking radiometer of PRISM L1B2 (left: X, and right: Y). Time trends of geometric correction accuracy of PRISM/N L1B2 using GCPs Absolute geometric correction (system correction) accuracy Period: June 4, 2007 June 3, 2008; 270 scenes, 3,354 GCPs as check points 04/01/08 06/01/08 08/01/

28 Geo Cal PRISM Geometric Correction Accuracy Averaged error in X (m) Averaged error in Y (m) /01/07 06/01/07 08/01/07 10/01/07 12/01/07 01/31/08 04/01/08 06/01/08 08/01/08 04/01/07 06/01/07 08/01/07 10/01/07 12/01/07 01/31/08 04/01/08 06/01/08 08/01/08 STDV of error in X (m) /01/07 06/01/07 08/01/07 10/01/07 Observation date 12/01/07 01/31/08 Observation date 04/01/08 06/01/08 08/01/ STDV of error in Y (m) /01/07 06/01/07 08/01/07 10/01/07 Observation date Averaged geometric errors of forward looking radiometer of PRISM L1B2 (left: X (pixel) direction, and right: Y (line) direction). 12/01/07 01/31/08 Observation date Standard deviations of geometric errors of forward looking radiometer of PRISM L1B2 (left: X, and right: Y). Time trends of geometric correction accuracy of PRISM/F L1B2 using GCPs Absolute geometric correction (system correction) accuracy Period: June 4, 2007 June 3, 2008; 177 scenes, 2,087 GCPs as check points 04/01/08 06/01/08 08/01/

29 Geo Cal PRISM Geometric Correction Accuracy Averaged error in X (m) Averaged error in Y (m) /01/07 06/01/07 08/01/07 10/01/07 12/01/07 01/31/08 04/01/08 06/01/08 08/01/08 04/01/07 06/01/07 08/01/07 10/01/07 12/01/07 01/31/08 04/01/08 06/01/08 08/01/08 STDV of error in X (m) Observation date STDV of error in Y (m) Observation date Averaged geometric errors of backward looking radiometer of PRISM L1B2 (left: X (pixel) direction, and right: Y (line) direction) /01/07 06/01/07 08/01/07 10/01/07 12/01/07 01/31/08 04/01/08 06/01/08 08/01/08 04/01/07 06/01/07 08/01/07 10/01/07 12/01/07 01/31/08 04/01/08 06/01/08 08/01/08 Observation date Observation date Standard deviations of geometric errors of backward looking radiometer of PRISM L1B2 (left: X, and right: Y). Time trends of geometric correction accuracy of PRISM/B L1B2 using GCPs Absolute geometric correction (system correction) accuracy Period: June 4, 2007 June 3, 2008; 212 scenes, 3,231 GCPs as check points

30 Image Quality PRISM S/N Evaluation S/ N Nadir Forward Backward Rad. Cal. JPEG filter before Processing Ver. Time trend of S/N evaluation for PRISM Signal-to-Noise ratio (S/N or SNR) is a parameter to show image quality S/N is also evaluating since launch the ALOS S/N increases by updating processing software due to reflect of radiometric calibration and JPEG filter implementation

31 Image Quality PRISM MTF Evaluation - Level 1B1 Nadir CCD1 CCD2 CCD3 CCD4 CCD5 CCD6 AT CT AT CT AT CT AT CT AT CT AT CT Forward CCD1 CCD2 CCD3 CCD4 AT CT AT CT AT CT AT CT CCD5 CCD6 CCD7 CCD8 AT CT AT CT AT CT AT CT Backward CCD1 CCD2 CCD3 CCD4 AT CT AT CT AT CT AT CT CCD5 CCD6 CCD7 CCD8 AT CT AT CT AT CT AT CT Level 1B2 Nadir 0.18 Forward 0.16 Backward 0.18 Modulation Transfer Function (MTF) is also evaluating since launch No significant change since launch the ALOS Around 0.2 in every condition : Specification is 0.2 CT > AT Forward have slightly smaller values 1B2 have slightly smaller values due to re-sampling

32 Image Quality AVNIR-2 MTF Evaluation - Level 1B1 0 deg. of pointing angle Band1 Band2 Band3 Band4 AT CT AT CT AT CT AT CT deg. of pointing angle Band1 Band2 Band3 Band4 AT CT AT CT AT CT AT CT Level 1B2 Band1 Band2 Band3 Band4 Y 方向 X 方向 Y 方向 X 方向 Y 方向 X 方向 Y 方向 X 方向 Modulation Transfer Function (MTF) is also evaluating since launch No significant change since launch the ALOS Around 0.21 to 0.30 (Specification Band 1-3 > 0.25, Band 4 > 0.2) CT > AT Large pointing angle degrade along track (AT) MTF of 1B1 1B2 have slightly smaller values

33 Geo Cal PRISM Alignment Model in EORC Exterior orientation parameters (Orbit and Attitude model) Orbit data: Determined by the onboard GPSR and ground processing (GUTS) Position errors are confirmed at less than 1m (3σ) by SLR validations. Attitude data: Satellite attitude is determined by the onboard STT (Star Tracker) and ground processing (PPDS) PRISM attitude are determined from the STT attitude and those alignment models on the optical bench. PRISM alignments on the optical bench (PRISM alignment) changes along with the thermal condition. The key to the precise attitude determination is the modeling of those alignment changes. STT Optical Bench x Flight direction z Earth center direction NDR BWD FWD PRISM and STT alignments on optical bench

34 Geo Cal PRISM Alignment Model in EORC Trend model of PRISM alignment changes To model Roll and Pitch Euler angle changing / Yaw is negligible Trend parameters: Short term alignment changing trend with orbit cycle S which corresponds to major changes of satellite thermal conditions Long term alignment changing trend with observation date D which corresponds to minor changes of satellite thermal conditions Exterior orientation Roll and Pitch bias errors of sample scenes which have various S and D are used for the trend modeling. Tsc -Tec Sun Scene cent er Tsc Earrh Center T Opening of satellite eclipse Tec Orbit cycle S = (0~1) T sc T T ec Orbit cycle parameter S

35 Geo Cal PRISM Alignment Model in EORC Roll Error [deg] Pitch Error [deg] s s Short term variation model as a function of s normalized observing time in recurrent (left: rolling, and right: pitching) Roll Error [deg] Pitch Error [deg] d d Long term variation model as a function of d observing date (left: rolling, and right: pitching). PRISM sensor alignment model (nadir) using GCPs Two time scale: short term - the time (2 nd order Fourier series), long term - the date (linear) Similar analysis for forward- and backward-looking radiometers Validation of pointing alignment parameter and generation of high level products

36 PRISM Calibration & Validation Plan Data flow Level 1 1A (Raw data) 1B1 (Radiometric correction) 1B2 (Geometric correction) Work flow Geometric Calibration Relative alignment between CCD Pointing/attitude Determination accuracy Stability Control accuracy Geometric correction accuracy Radiometric model Radiometric Calibration #1 Evaluation of cal. mode data Stripe Linearity Offset and slope Relative accuracy between optics PPDS (attitude/pointing determination) Establishing ground control point (GCP) Rational Polynomial Camera (RPC) model Absolute orientation Relative orientation Geometric model Stereo pair Matching Image Quality Evaluation MTF, S/N Lossy, JPEG extension Matching accuracy Co-registration with AVNIR-2 Radiometric Cal. #2 Absolute calibration Aim to geometric accuracy of 1B2 Validation Absolute DSM Relative DSM Ortho Image Aim to radiometric accuracy of 1B2 Cooperation with; AVNIR-2 for radiometric calibration and geometric calibration at nadir, and Precision Pointing and Geolocation Determination System (PPDS) for satellite position and attitude evaluations : GPSR, STT, IRU, and ADS