In-Flight PSF/MTF analysis of Landsat 8 Thermal Infrared Sensor (TIRS) CEOS/IVOS MTF Workshop ONERA, Toulouse, France November 16, 2015 Dennis Helder

Transcription

1 In-Flight PSF/MTF analysis of Landsat 8 Thermal Infrared Sensor (TIRS) CEOS/IVOS MTF Workshop ONERA, Toulouse, France November 16, 2015 Dennis Helder

2 Objective Site Selection Methodology Results/Analysis Conclusions Outline YEMEN Much of the work was performed by Jason Hong Acknowledgement: This work was supported by the NASA Landsat Project Science Office and USGS EROS. Pre- and Post-Launch Spatial Quality of the Landsat 8 Thermal Infrared Sensor Brian N. Wenny, Dennis Helder, Jungseok Hong, Larry Leigh, Kurtis J. Thome and Dennis Reuter Remote Sens. 2015, 7(2),

3 Objectives On-orbit estimation of PSF/MTF for thermal sensors is typically difficult to do and not commonly performed due to difficulty in target development and spatial resolution. Can natural edge targets be located for moderate resolution thermal sensors? Large temperature transitions often occur at land/water interfaces cold ocean water next to warm land surfaces. Likely candidates are in northern Africa and Middle East.

4 Required conditions Site Selection- MTF 1. Edge : Coastal line, land-water 2. Angle appropriate to sensor orbital path --typically 6-8 o difference 3. Maximal temperature difference 4. Minimal interfacial region extent 5. Uniform temperature on either side of edge YEMEN YEMEN

5 MTF Site Selection

6 Site Selection- MTF

7 MTF Analysis : OMAN Oman P158 R46 OMAN OMAN OMAN

8 MTF Analysis : YEMEN Yemen P163 R50 YEMEN YEMEN YEMEN

9 MTF Analysis : WESTERN SAHARA 1 Western Sahara P205 R42 Western Sahara Western Sahara 77.65m Western Sahara

10 MTF Analysis : WESTERN SAHARA 2 Western Sahara P206 R45 Western Sahara Western Sahara Western Sahara 55m

11 MTF Analysis : WESTERN SAHARA 3 Western Sahara P206 R45 Western Sahara Western Sahara Western Sahara 61m

12 Landsat 8 TIRS Images of Test Sites Site Number of Scenes Start Date End Date Oman Apr May-2014 Yemen 5 27-Oct Apr-2014 Western Sahara June Jun-2014 Western Sahara May Apr-2014 Western Sahara May Apr-2014 Cross-Track Along-Track Cross-Track Cross-Track Cross-Track

13 Methodology 1. Locate edge to sub-pixel accuracy by fitting Fermi function Fermi function modified to account for temperature gradient 2. Align edge slices 3. Smooth with modified Savitzky-Golay filter 4. Differentiate to obtain PSF 5. Fourier transform, and normalize to obtain MTF

14 Procedure plots for edge method Pixel Angle=5.404 [deg] Test image Sub-pixel edge location Least square error line Pixel DN msg filtering with raw data Raw data msg filtering DN diff = STD ave = SNR = Pixel Normalized Differentiation value Trimmed LSF FWHM = [Pixel] Pixel Normalized value MTF Value at Nyquist = Normalized frequency Stennis tarp target MTF result of IKONOS on 8/1/2005 with MTFC off. 14

15 Parametric Edge Detection A model-based parametric method was applied to estimate edge location to sub-pixel accuracy. The Fermi function was chosen as a parametric fit to locate the edge to sub-pixel accuracy. Parameter e is the sub-pixel edge location estimate. Linear term added to account for colder water and warmer land farther away from edge f ( x) b d d s( x 1 e e) gx 15

16 Modified Savitzky-Golay (msg) Filtering Unlike normal Savitzky-Golay filtering, msg filter is applicable to nonuniformly spaced data. Using a least-squares approach, a 3 th order polynomial was fitted to the data in a 1.3-pixel wide window centered on each location where an output value was desired. 16

17 One output in the window center is evaluated by the polynomial. The next value is evaluated by shifting the window at a sub-pixel scale. (typically 0.05 GSD) The shifting step determines output resolution. MSG filtering 17

18 Example of Edge Profile f ( x) b d d s( x 1 e e) gx ESF example illustrating linear temperature gradient and variable land temperature.

19 Results OMAN YEMEN

20 Results SAHARA 2 SAHARA 3

21 Summary RER and FWHM Data Sites Orientation Path/Row SNR RER STD FWHM (30m pixel) STD FWHM (meters) STD Oman Across 158/ Yemen Along 163/ Western Sahara 1 Across 205/ Western Sahara 2 Across 206/ Western Sahara 3 Across 206/

22 Conclusions Natural water/land edges exist that make it possible to estimate PSF/MTF for moderate spatial resolution (~100m GSD) thermal sensors. Consistent results can be obtained for common metrics such as RER and FWHM with uncertainties on the order of 2-4% Yemen and Sahara 3 test sites exhibit best consistency and form a basis for along-track and across-track estimation. Temperature changes on either side of the edge may be modeled. However contribution to PSF/MTF estimates due to a temperature transition region at the edge is not precisely known, but thought to be small with judicious target selection. Results are consistent with pre-launch estimates but difficult to compare due to ground processing of the post-launch data to a level L1T terrain corrected product.