36
ERDC TR-05-1
cross-track direction. These are the outside pixels along the northern and south-
ern edge of each flightline.
Several methods were employed to deal with the negative DNs. If all the
anomalous pixels were indeed located at the edge of the flightline, they were
simply assigned a value of 0, becoming background pixels. In this case, all 38
bands lost these edge pixels to maintain the proper file size of the actual image
data. If one or more of the negative pixels within an individual band were located
in the interior of the image data, they were renumbered to a value of 1. This time
consuming step in the post-processing the AISA data had very little impact on
the overall radiometric quality of the imagery. However, it was thought to be
critical in preparing the images for mosaicking. No other manual adjustments to
the image histograms were employed.
3.3.2.2 Specular Reflectance or Water Glint. The saturated, or washed-out,
water pixels are a normal phenomenon in both digital and analogue images. The
AISA images are no exception. The saturation is a result of the specular reflec-
tance properties of water. Specular reflectance occurs when nearly all of the in-
coming illumination (in this case, sunlight) is reflected from a non-diffuse, or
smooth, surface (such as calm water) directly onto the optical focal plane. In
other words, the water acts like a mirror, with the reflected sunlight saturating
these pixels. Figures 18 and 19 clearly depict these saturated water pixels. Also,
note that the glare decreases in intensity from south to north across the image.
The glinted pixels diminish towards the north side of the flightline because the
mirror (i.e., the water) reflects the sunlight at an angle that passes below the far
end of the detector. Any subsequent radiometric manipulations to try to improve
the quality of the vegetated pixels would likely be negatively affected by the
presence of the glaring water pixels. Therefore, a masking procedure was im-
plemented to remove all the pixels containing standing water.
Several techniques were considered for removing the water pixels. Both su-
pervised and unsupervised classification routines, applied to each flightline inde-
pendently, were investigated, but were determined to require too much time and
employed subjective delineation of land cover classes. Digitizing the water areas
was also considered, but again would have required too much effort. The
land/water mask for each flightpath was eventually created using the Normalized
Difference Vegetation Index (NDVI). This simple mathematical expression uses
the ratio of the difference between the reflectance DNs of a near infrared band
and a red band as compared to their sum.
Previous Page
