Figure 1.
A subset image extracted from hyperspectral data acquired over the test site on 3 September 2011 at 06∶42 UTC.
The locations of the 23 targets (R1–R4, H1–H4, and M1–M15) are displayed in the image. A bare area highlighted in a white rectangle is used to perform the signal-to-noise ratio estimation. The two pixels labeled as P1 and P2 are used to demonstrate the discrepancy between the uniform and non-uniform Lambertian land surface reflectance.
Figure 2.
Center wavelengths of 128 bands of the UAV-HYPER sensor.
Table 1.
Main characteristics of the UAV-HYPER sensor.
Figure 3.
In situ surface reflectance spectra of the 23 targets (R1–R4, H1–H4, and M1–M15) in the wavelength range of 0.4–0.95 µm.
The four dashed vertical lines denote the positions of the center wavelengths of four selected bands (19, 33, 54, and 83) that represent the blue, green, red, and near-infrared bands, respectively.
Figure 4.
Signal-to-noise ratio estimated using the bare area shown in Figure 1 and the target area including targets R1–R4 and H1–H4.
The bands with signal-to-noise ratio values lower than 40 are discarded in this study.
Figure 5.
Flowchart of the radiometric calibration procedure of the UAV-HYPER sensor.
Figure 6.
At-sensor radiances Lsensor as a function of the DN for bands 19, 33, 54, and 83.
is the center wavelength.
Figure 7.
Comparison of the apparent reflectance with the simulated at-sensor reflectance
for targets H1–H4 and M1–M15 in bands 13–108.
Different symbols with different colors represent different targets.
Figure 8.
Root mean square error (RMSE) and relative RMSE (RRMSE) values of apparent reflectance as a function of the wavelength for targets H1–H4 and M1–M15.
A: RMSE and RRMSE values between the apparent reflectance and the simulated at-sensor reflectance
for targets H1–H4 and M1–M15 in bands 13–108. B: Same as Figure 8A, but the apparent reflectance greater than 0.45 in bands 13–108 has been discarded due to the non-linear response of the UAV-HYPER sensor.
Table 2.
Breakpoint positions in the atmospheric LUT for the six input parameters.
Figure 9.
Flowchart of the atmospheric correction procedure for the UAV-HYPER data.
Part 1 is used to derive the uniform Lambertian surface reflectance using the atmospheric look-up table (LUT). Part 2 is used to determine the window size (WS) of the atmospheric point spread function (PSF) by calculating the
difference between two successive iterations. Part 3 is used to determine the final non-uniform Lambertian surface reflectance
by calculating the
difference between two successive iterations.
Figure 10.
Comparison of the uniform Lambertian surface reflectance derived using Equation (5) with the apparent reflectance
and the in situ surface reflectance
for targets H1–H4 in bands 13–108.
Figure 11.
Comparison of the uniform Lambertian surface reflectance derived using Equation (5) with the in situ surface reflectance
for targets H1–H4 and M1–M15 in bands 19, 33, 54, and 83.
is the center wavelength of each of the four bands.
Figure 12.
Comparison of the uniform Lambertian surface reflectance derived from Equation (5) and the in situ surface reflectance
for targets H1–H4 and M1–M15 in bands 13–108.
Different symbols with different colors represent different targets.
Figure 13.
Relative errors of surface reflectance as a function of the wavelength for targets H1–H4.
Relative errors between the uniform Lambertian surface reflectance derived from Equation (5) and the non-uniform Lambertian surface reflectance
derived from Equation (7) and the in situ surface reflectance
for targets H1–H4 in bands 13–108.
Figure 14.
Comparison of the uniform Lambertian surface reflectance derived from Equation (5) and the non-uniform Lambertian surface reflectance
derived from Equation (7) for pixels (A) P1 and (B) P2.
is the surface reflectance difference between
and
(
).
Figure 15.
Surface reflectance differences (
) as a function of the wavelength for the four AOD@550 values for pixels (A) P1 and (B) P2.
Figure 16.
Root mean square error (RMSE) and relative RMSE (RRMSE) values of surface reflectance as a function of the wavelength for targets H1–H4 and M1–M15.
A: RMSE and RRMSE values between the non-uniform Lambertian surface reflectance derived using Equation (7) and the in situ surface reflectance
for targets H1–H4 and M1–M15 in bands 13–108. B: Same as Figure 16A, but surface reflectance greater than 0.5 in bands 13–108 is excluded.