Page 10 - Forum-2021-JanToMarch

reconstructed distance to the object:
Figure 4. Surveillance camera video.
Figure 6. Camera location shown in Google Earth.
Figure 7. Lines of sight of witness statement (left) and reconstructed location with
error (right).
Figure 8. Single frame of video showing the B-767 behind trees (left) and with
overlay of 3-D object (right).
10 • January-March 2021 ISASI Forum
The Evaluation Case
On Feb. 23, 2019, the Amazon Prime Air cargo
aircraft that was operated by Atlas Air flew
from Miami, Florida, to Houston, Texas, and
during the arrival phase the Boeing 767-375BCF
entered a rapid descent from 6,200 feet and im-
pacted into a marshy bay area around 40 miles
away from Houston’s George Bush Interconti-
nental Airport (see Figure 3, page 9).
Two security cameras captured the last 5
seconds of the aircraft in a steep, generally
wings-level attitude until impact with the
swamp. For our own interest, the iwi® method-
ology was applied to validate the latest meth-
odology developments, reconstructing flight
Figure 5. Surveillance camera video. path, aircraft attitude, and ground speed based
on the available video information. The results
were shared with the NTSB before the black box
could be recovered.
Flight Data Reconstruction and Data Com-
parison
Two video recordings were available from two
different locations. They were used to create
panoramic images with aircraft positions and
video time stamps. The video distortion was
corrected using the dedicated lens correction
profile (see Figures 4 and 5).
The resolution of the video influences the
data reconstruction accuracy as well as the
precision of the time stamp information. Video
frames at every second were selected to ensure
maximum accuracy. The security camera re-
cording frequency was approximately one frame
per second.
The GPS location of both cameras was iden-
tified, as shown in Figure 6, using Google Street
View, as well as the location of reference objects
like buildings and trees, as shown in the video.
The location accuracy was defined with +/- 16
feet in latitude/longitude and altitude. The
reference objects were used to place the image
information within the 3-D environment. All
data collected was imported into the recon-
struction software called Immersive Witness
Analyzer, which sets all lines of sight consider-
ing reference objects. The software estimates
the reconstructed flight path considering
potential error information using the specified
and decribed formulas.
To estimate the aircraft’s attitude at different
locations along the flight path, a 3-D model of a
B-767 was placed and adjusted with respect to
the video image frame rate.
The left image in Figure 7 shows the lines
of sight from the perspectives of two security
cameras. The intersection of both lines ap-