The imagery was collected using ADS40-SH51 and ADS40-SH52
digital sensors. Collection was performed using a
combination of twin-engine aircraft flying at an average
flying height of 17500 ft and 27000 ft above mean terrain
respectively, with nominal sidelap of 23%, giving the
collected data nominal ground sampling distance of 0.66
meters and 0.9 meters, respectively. Based-upon the CCD
Array configuration present in the ADS40 digital sensor,
imagery for each flight line is 12,000-pixels in width.
Red, Green, Blue, Near-Infrared and Panchromatic image
bands were collected.
Collected data was downloaded to portable hard drives and
shipped to the processing facility daily. Raw flight data
was extracted from external data drives using GPro software.
Airborne GPS / IMU data was post-processed using IPAS,
PosPac and/or TerraPos software and reviewed to ensure
sufficient accuracy for project requirements.
Using Pictovera software, planar rectified images were
generated from the collected data for use in image quality
review. The planar rectified images were generated at
five meter resolution using a two standard deviation
histogram stretch. Factors considered during this review
included but were not limited to the presence of smoke
and/or cloud cover, contrails, light conditions, sun
glint and any sensor or hardware-related issues that
potentially could result in faulty data. When necessary,
image strips identified as not meeting image quality
specifications were re-flown to obtain suitable imagery.
Aerotriangulation blocks were defined primarily by order
of acquisition and consisted of four to seventeen strips.
Image tie points providing the observations for the least
squares bundle adjustment were selected from the images
using an autocorrelation algorithm. Photogrammetric control
points consisted of photo identifiable control points,
collected using GPS field survey techniques. The control
points were loaded in to a softcopy workstation and
measured in the acquired image strips. A least squares
bundle adjustment of image pass points, control points
and the ABGPS was performed to develop an aerotriangulation
solution for each block using Pictovera software. Upon
final bundle adjustment, the triangulated strips were
ortho-rectified to the USGS NED DEM for the project area.
A combination of 10-Meter and 30-Meter NED data purchased
from USGS in 2005 was used for rectification. The images
were re-sampled from the raw resolution of 0.7 meters to
the required resolution of 1.0 meters.
Positional accuracy was reviewed in the rectified imagery
by visually verifying the horizontal positioning of the
known photo-identifiable survey locations using ArcGIS software.
The red, green, blue, and NIR bands were combined to
generate a final ortho-rectified image strip. The ADS40
sensor collects twelve bit image data which requires
radiometric adjustment for output in standard eight bit
image channels. The ortho-rectified image strips were
produced with the full 12 bit data range, allowing
radiometric adjustment to 8 bit range to be performed on
a strip by strip basis during the final mosaicking steps.
The imagery was mosaicked using manual seamline generation
in Orthovista Seam Editor (OVSE). The 12 bit data range
was adjusted for display in standard eight bit image
channels by defining a piecewise histogram stretch using
OrthoVista software. A constant stretch was defined for
each image collection period, and then strip by strip
adjustments were made as needed to account for changes
in sun angle and azimuth during the collection period.
Strip adjustments were also made to match the strips
histograms as closely as possible to APFO specified
histogram metrics and color balance requirements.
Automated balancing algorithms were applied to account
for bi-directional reflectance as a final step before
the conversion to 8 bit data range.
APFO specified DOQQs were extracted from the final mosaic
in GeoTIFF format. 4-Band DOQQs were produced and 3-Band
RGB CCMs were created. DOQQs corresponding to an
individual CCM were reviewed for overall color balance
within the CCM. Local corrections were made where
necessary to ensure uniformity within the CCM. In the
case of DOQQs occurring in more than one CCM, a separate
version of the image was generated and balanced for
each CCM it occurred in.