36th CEOS Working Group on Calibration and Validation Plenary May 13-17, 2013 at Shanghai, China UAV-based Remote Sensing Payload Comprehensive Validation System Chuan-rong LI Project PI www.aoe.cas.cn Academy of Opto-Electronics (AOE), Chinese Academy of Sciences (CAS) 15 May 2013
Outline 1 Background 2 3 4 System Overview Inflight Calibration & Performance Assessment Future Plan 2
1. Background Trend of Earth observation systems: Quantitative remote sensing application International standardization of data quality assessment and data sharing International EOS Pre-launch Cal &Val demand Monitoring of payload performance in system operation period 3 3
1. Background LOW COST Development cost:10% of manned aircraft Operation cost:10% of manned aircraft Provide realistic validation environments (telemetry, telecontrol, etc) Carry out flight experiment in specific scenarios UAV-based Remote Sensing Payload Comprehensive Validation System 4
2. System Overview Two UAV Platforms Optical and SAR RS sensors System Components Comprehensive Cal&Val sites Data processing/analyzing systems Auxiliary support systems 5
2. System Overview - The Comprehensive C&V Site Two comprehensive Cal&Val test sites were preliminarily established, which have different climate characteristics, various land cover types and topographic features. Now they are being further developed so as to support multi-grade validation of airborne and spaceborne sensors for stable and long-term operation. North China site South China site 6
2. System Overview - The Comprehensive C&V Site Standard artificial and natural targets Knife-edge target Fan-shaped target Trihedral Corner Reflector Dihedral Corner Reflectors Rice Maize Gray-scale target Colored target Ku-band Active Transponder Potato Sunflower Three-bar target Layout of targets Surface parameters measurement Spectral curves Optical targets SAR targets Natural ground targets 7
2. System Overview - Data processing/analyzing system Data processing/analyzing system Data processing Radiometric performance assessment Geometric Performance Assessment Spectral Performance Assessment Radiometric correction Geometric correction Spectral correction Reflectance retrieval Vegetation index retrieval SAR data processing Absolute radiometric calibration Signal to Noise Ratio, SNR Dynamic range Response linear degree Radiometric resolution, NE ρ Ground resolution MTF Band registration precision Hyperspectral camera: central bandwidth; FWHM Multispectral camera: spectral response function 8
3. Inflight Calibration & Performance Assessment Flight campaigns Nov 2010 Campaign Airborne optical sensors in North China test site Jul 2011 Campaign Airborne optical and SAR sensors in South China test site Sep 2011 Campaign Airborne optical and SAR sensors in North China test site 9
0.080 0.060 0.040 0.020 3. Inflight Calibration & Performance Assessment Atmospheric and field measurements Spectral reflectance of targets were measured. The aerosol optical thickness data and meteorological profile above the test site were synchronously collected. 0.000 09:11:10 11:11:06 13:11:04 15:11:06 550nm aerosol optical thickness BRF 1.0 0.9 0.8 0.7 0.6 0.5 0.4 0.3 0.2 0.1 0.0 10 20 30 40 50 60 70 80 角度 ( ) MFT-60 450nm MFT-60 650nm MFT-60 900nm MFT-30 450nm MFT-30 650nm MFT-30 900nm MFT-20 450nm MFT-20 650nm MFT-20 900nm BRF properties of targets Spectral Target reflectance reflectance 实测靶标光谱反射率 by of filed gray-scale measurement targets Surface reflectance 60% 50% 40% 30% 20% 10% 0% 0.4 0.5 0.6 0.7 0.8 0.9 1.0 Wave length(um) 50% 40% 30% 20% 0.15 0.10 Radiosounding balloon 0.05 0.00 09:11:10 11:11:06 13:11:04 15:11:06 Moisture content Automatic sun tracking photometer, CE318 Automatic weather station 10
3. Inflight Calibration & Performance Assessment Cal&Val of Optical sensors Absolute radiometric calibration Gain 0.7 0.6 0.5 0.4 0.3 0.2 0.1 Hyperspectral The inflight calibration coefficient The calibration coefficient at laboratory 0 20 40 60 80 100 120 Channel Multispectral Flow chart of optical sensor radiometric calibration Radiometric calibration coefficients have very good linearity and the correlation coefficient reaches above 99%. 11
3. Inflight Calibration & Performance Assessment Cal&Val of Optical sensors Relative radiometric calibration Because of the variation in velocity height ratio, the images of hyperspectral imager between adjacent flight strips lack of comparability for different surfaces. A relative radiometric calibration method based on line frequency difference is proposed to solve this problem. 12
Real central wavelength Central wavelength 3. Inflight Calibration & Performance Assessment Cal&Val of Optical sensors Spectral calibration 770 769 768 767 766 765 764 763 762 761 中心波长 Spectral calibration for hyperspectral sensor 280~315 line 768 766 764 762 760 758 756 754 1 101 201 301 401 501 601 701 801 901 1001 Pixel line 1 46 91 136 181 226 271 316 361 406 451 496 541 586 631 676 721 766 811 856 901 946 991 Pixel line Retrieval central wavelength based on flight data in 2011. Band 75 is O 2 absorption band. (laboratory measurement of Band 75 in Sep. 2010) 12 10 8 6 4 2 0 (2) Shift of central wavelength in 2011 1 101 201 301 401 501 601 701 801 901 1001 The shift of central wavelength is approximately 4~6nm compared to laboratory measurement. z 13
3. Inflight Calibration & Performance Assessment Cal&Val of Optical sensors Spectral calibration Spectral calibration for multispectral sensor Difficulty: Solving of spectral response function faces ill-condition matrix Solutions: The spectral reflectance of 15 multispectral targets were measured to add the number of equations; Piecewise fitting SRF according to laboratory measurements Laboratory Retrieval Response rate Response rate wavelength (nm) Response rate Response rate wavelength (nm) wavelength (nm) 14
3. Inflight Calibration & Performance Assessment Cal&Val of Optical sensors True-color calibration A general true color calibration model is proposed based on the physical mechanism of color generation, which can fully employed all spectral information in VNIR reflection zones of hyperspectral images. In-situ measured spectral reflectance of the targets 1 At-sensor ideal true color of the targets 3 2.5635-0.3903-0.2178 M RGB > R' G ' B' = -0.1077 3.7997 0.5655 0.0416 0.2628 12.4262 MODTRAN 2 True color correction coefficient matrix 4 Human visual color matching function Reconstructed true color of the targets & Images before true color correction 4 At-sensor radiance HSI image True color correction The corrected true-colored image
3. Inflight Calibration & Performance Assessment Cal&Val of Optical sensors Geometric performance Ground resolution Defined as the least ground distance or the least size of object that can be distinguished. Image of radiometric target Extracte the grey value from the circumference of radius equal to r Calculate the contrast of this radius Distinguish? No r = r+1 Yes Sensor GSD Panchromatic image in 2011 Hyperspectral image in 2011 Red line denotes the location of ground resolution estimated by our method. Blue line denotes the location of GSD. Green line denotes the location of ground resolution estimated by visual method. Camera Calculated Visual resolution Resolution(m) (m) GSD(m) Panchromatic 0.7941 0.8153 0.35 B1 0.8889 0.8178 0.70 Multispectral B2 0.7800 0.8083 0.70 B3 0.8875 0.8428 0.70 B4 0.7500 0.7750 0.70 16
3. Inflight Calibration & Performance Assessment Cal&Val of Optical sensors Geometric performance Geometric distortion Geometric distortion evaluation result of panchromatic image in Sep. 2011 in Baotou city, China. Geometric distortion evaluation module External distortion Positioning error Internal distortion Data Type X-direction (m) Y-direction (m) Distance distortion (m) Angle distortion ( ) Geometric coarse correction Geometric precision correction 22.106 155.295 14.859 167.16 0.568 1.012 0.866 0.085 After geometric precision correction, the positioning accuracy can reach a meter-scale and the image distortion has also been corrected well. 17
3. Inflight Calibration & Performance Assessment Cal&Val of SAR sensors SAR data processing system Raw Data Motion Compensation High resolution imaging processing Radiometric and geometric processing Interferometric Processing and DEM generation Polarimetric SAR data Processing and correction 18
3. Inflight Calibration & Performance Assessment Cal&Val of SAR sensors Relative radiometric calibration Before relative radiometric calibration After relative radiometric calibration 19
3. Inflight Calibration & Performance Assessment Cal&Val of SAR sensors data analyzing system Point target analysis DEM analysis Resolution PSLR(Peak Side Lobe Ratio) ISLR(Integrated Side Lobe Ratio) Dynamic range. N o. Az. Res. (m) Rg. Res. (m) Az. PSLR (db) Rg. PSLR (db) Az. ISLR (db) Rg. ISLR (db) Dy. Range (db) 1 0.2997 0.2707-29.9001-24.3671-30.5193-25.5551 85.8664 2 0.2997 0.2804-31.7376-21.9761-29.8146-24.7290 85.2708 3 0.3045 0.2804-29.2417-22.0863-29.5731-25.9416 86.4129 4 0.2997 0.2852-30.0977-32.3270-28.4382-25.0391 82.5886 5 0.2997 0.2852-31.9878-21.1309-29.6991-19.2911 85.4210 6 0.3190 0.2852-27.9996-26.5897-28.1798-27.0331 77.4594 7 0.3094 0.2900-30.4414-23.3795-29.8154-26.2518 86.1319 8 0.2997 0.2755-28.2101-24.5004-30.2793-25.8863 87.7243 9 0.2997 0.2804-28.7574-26.5284-29.1662-27.6142 84.7408 1 0 0.3094 0.2852-28.8182-22.7427-29.9757-24.0932 81.8551 DEM accuracy ENL Radiometric resolution Distributed target analysis 20
3. Inflight Calibration & Performance Assessment Application performance reflectance retrieval and validation A Look Up Table (LUT) atmospheric correction model with adjacency effect correction was proposed to retrieve land surface reflectance. Target 1 Target 2 Target 3 Target 4 Colored targets in red frame ref1: at-sensor reflectance without atmospheric correction ref2: surface reflectance after atmospheric correction ref3: field measured reflectance It can be seen that the proposed method can eliminate the atmospheric effect well. 21
3. Inflight Calibration & Performance Assessment Application performance LAI retrieval and validation Leaf Area Index (LAI) was retrieved from hyperspectral data according to the image classification. Retrieve (1) Accuracy of LAI retrieval model is less than 7%. Airborne hyperspectral reflectance three-dimensional cube 5 Result of LAI retrieved (2) Validation results show that the retrieval error of LAI is approximately 21.7% with field measurement data. Accuracy of LAI retrieval model Six vegetation types Accuracy of LAI retrieval model for six vegetation types LAI Measured 4.5 4 3.5 3 2.5 2 1.5 1 0.5 0 0 1 2 3 4 5 LAI Retrieved from Reflectance Measured Accuracy assessment using field measurements on Sep 3, 2011. (3) It might due to the retrieval error of reflectance, the saturation of NIR band with the increasing of LAI, the error of instruments. 22
4. Future Plan satellite Dome C Data process vehicle Command vehicle Airborne platform Baotou Comprehensive Cal&Val Site Aircraft Hangar Equipment store house Tower T&C vehicle Real-time monitoring of flight status 1 0.35 3 0.75 Artificial target Natural scene 0.4 0.8 4 Inflight and preflight geometric calibration 23
www.aoe.cas.cn