The Terminal Radar Approach Control (TRACON) air traffic controller is
responsible for the safety of arriving and departing aircraft within the
terminal area. The air traffic
controller must ensure aircraft are spaced no closer than 1,000 vertical feet
and 3, 4, or 5 lateral separation, depending upon aircraft size (Wickens,
Mavor, and McGee, 1997). For many
years, the Federal Aviation Administration (FAA) has introduced new technologies
and procedures to be integrated into the National Airspace System (NAS) to
improve air traffic controller performance and minimize users’ risk. One such program, called the Integrated Terminal Weather System (ITWS),
provides TRACON controllers short-term forecasting of hazardous weather patterns. The purpose of this study is to develop a
color discrimination model to assist human factors professionals in determining
the optimal assignment of look-up-table values to ITWS objects overlaid on the
TRACON radar display.
The ITWS
integrates FAA and National Weather Service sensors to provide TRACON
controllers a weather model that can predict hazardous weather conditions 30
minutes into the future (Cole and Wilson, 1994). This weather model can predict storm motion, leading edge of
storms, gust fronts, microbursts, windshear, tornado, wind speed, and lighting
strikes. The ITWS data will be overlaid
on the Standard Terminal Automatic Replacement System (STARS) or the Automated
Radar Terminal System (ARTS) color display.
The depiction of weather symbols size or color has not been determined
nor has the FAA finalized the procedures on how the TRACON controller will
utilize this information. Currently,
the air traffic controller, TRACON or en-route, does not separate aircraft from
weather, but they can and should provide weather advisories to flight
crews. The flight crew may request a
heading or altitude change based upon the weather advisory, but it is the
responsibility of the flight crew to avoid hazardous weather.
The TRACON radar display contains aircraft data blocks
(aircraft’s call sign, mode C altitude for equipped aircraft, and ground speed) and terminal area markers
such as ground hazards, approach and departure routes, and navigational fixes (Wickens, Mavor, and
McGee, 1997). Currently, controllers
use the Full Digital ARTS Display (FDAD) monochrome monitor, but the FAA is
replacing these monitors with the STARS or ARTS color displays. Both the STARS and ARTS color display can
support ITWS software, but there is no consensus as to the best method to
overlay weather information on the radar display.
Scharff, Hill, and Ahumada (2000) found that the readability of
text on textured backgrounds can be predicted by a simple luminance masking
metric, which computes an equivalent contrast on a uniform background. The metric was borrowed from contrast gain
control masking models and has the form
CE = C/(1 + (CRMS/C2)2)0.5
,
where CE is the
equivalent text contrast, C is the text luminance contrast, CRMS is
the root mean square luminance contrast of the background alone and C2
is a luminance contrast masking threshold.
Our hypothesis is that this equivalent luminance contrast
metric will predict the equivalent search accuracy and latency performance in
our text block identification task.
METHODS
Observers:
Three observers (26, 35, and 37 years of age) had normal or
corrected-to-normal visual acuity, and two of the three observers had normal
color vision as tested with pseudo-isochromatic plates. Informed consent
was obtained from all observers. All
observers but one (first author who was red/green color deficient) were naïve
to the experimental hypothesis.
Apparatus:
Stimuli were displayed on a 19” color CRT monitor at a frame
rate of 85.0 Hz. Observers viewed the
screen from a distance of approximately 0.75 meters, giving 53.8 pixels per
degree of visual angle in the horizontal direction and 45.7 pixels per degree
vertically. The screen was 26.2 by 20.4
degrees (1024 by 768 pixels).
Stimuli: The stimuli were presented using the red, green, and blue guns of a color CRT. At their maximum level, the individual guns had the luminance (Y) and CIE color chromaticities (x, y) shown in Table 1.
|
Gun |
Y |
X |
y |
|
Red |
13.5 |
0.603 |
0.356 |
|
Green |
43.1 |
0.282 |
0.612 |
|
Blue |
5.22 |
0.143 |
0.063 |
Table 1. Luminance (cd/m2) and
chromaticity values for the three CRT guns.
The screen background was set
to a neutral gray with equal contributions (128/255) from the three guns.
Calibrations were done with a Minolta CS-100 colorimeter. Each image had fewer than 20 colors, so each
of the colors in the image and the grays used in the text were calibrated individually.
The
background images were 20.05 by 15.58 degrees (776 by 593 pixels). The uniform background had a luminance of
4.44 cd/m2. The wind shear
and gust front had mean luminances of 3.2 and 0.83 cd/m2,
respectively. The root mean square
contrasts for the map backgrounds were 1.53 and 3.56, respectively. The text contrasts relative to the
background mean luminance were: 0.11, 0.23, 0.55, and 0.86 for uniform; 1.03,
1.71, for wind shear; and 2.46, 4.58 for gust front.
Adobe
PhotoshopÒ version 5.5 was used to
generate the data blocks. The data
block image font was Letter Gothic MT size 10 points with tracking set at –70.
Figure 1. Uniform noise pattern. The test stimulus consisted of randomly
selected aircraft data blocks positioned in one of eight locations 1200, 0130,
0300, 0430, 0600, 0730, 0900, and 1030 respectively at a distance of 5.8
degrees from the center. The subject’s
task was to identify the position of the surround data block to the center data
block.
Figure 2. A simulated wind shear
weather pattern overlaid on an air traffic control radar monitor display. The test stimulus similar to the uniform
noise pattern consisted of randomly selected aircraft data blocks positioned in
one of eight locations 1200, 0130, 0300, 0430, 0600, 0730, 0900, and 1030
respectively at a distance of 5.8 degrees from the center. The subject’s task was to identify the
position of the surround data block to the center data block.
Figure 3.
A simulated gust front weather pattern overlaid on an air traffic
control radar monitor display.
Procedure: The
observers’ task was to search a simulated air traffic control weather radar
image for a specific aircraft data block. Each trial began with a fixation
cross located in the center of the background image. After 500 milliseconds, the fixation cross was replaced by a
stimulus image which remained visible for 8.5 seconds or until the observer’s
response. The stimulus image contained
a data block located in the center of the screen with an additional eight data
blocks positioned 5.8 degrees from the center.
The observer’s task was to provide a manual response indicating the
location of the surrounding data block that matched the center data block. Observers were asked to press ‘1', ‘2’, ‘3’,
‘4’, ‘6’, ‘7’, ‘8’, ‘9’ on the numeric keypad of a standard PC keyboard if a
target was located in the ‘0730’, ‘0600’, ‘0430’, ‘0900’, ‘0300’, ‘1030’,
‘1200’, and ‘0130’ respectively. The
participant began the next trial by pressing ‘5’. For each condition the background scene – either the uniform,
wind, or gust image without text blocks – remained visible throughout the 60
trials thereby avoiding any secondary masking effects due to switching between
no scene to scene for each trial. Error
rates and reaction times (RTs) were recorded.
Observers were asked to provide responses as quickly as possible while
maintaining a high level of accuracy.
Each observer completed 3 replications of the experiment. Each
replication consisted of eight blocks ( 4 text contrasts on the uniform
background, 2 text contrasts with the wind shear mask, and 2 text contrasts
with the gust front mask) of 20 trials each, for a grand total of 480
trials. Each subject received a
different order of blocks. Data block
location was not a factor, thus on some blocks target location was not
uniformly distributed across location.
Feedback was provided following an incorrect response. Individual trials were separated by
intervals of approximately 1000 milliseconds.
Observers were allowed periodic rest throughout the experimental
session.
RESULTS
Figure 4 shows the average proportion correct identifications
for the 3 observers a function of the text luminance contrast adjusted by the
masking metric with a masking threshold of C2 = 0.5 for each of the three backgrounds. Figure 5 shows individual data for the three
subjects which illustrates the same pattern as the average data. Figure 6 shows the average response times
in the same format and figure 7 displays each subject’s data which shows a
similar pattern to the average data.
The accuracy appears to have reached asymptote for a contrast of 0.23 on
the uniform background, but the response latencies continue to improve with
contrast. A masking threshold was
estimated for each of the four masked conditions separately from the response
latency data. These predictions were
obtained using linear interpolation on the latency data to find the uniform
background contrast that would give the latency obtained with the mask
background. The threshold of 0.5 is the
average of the four estimates. The
metric with this threshold gives an equivalent
contrast that is too low for the wind shear mask conditions and too high
for the gust front mask conditions.
Figure
4. Mean accuracy for three
subjects. The wind shear predicted mask
values were nearly identical to the behavioral wind shear mask values, while
the gust front predicted values were slightly different than the behavioral
values.
Figure 5. Three subjects’ individual accuracy performance results for the uniform fixed pattern, wind shear and gust front conditions.
Figure 6.
Mean reaction time for three subjects.
The gust front predicted mask values were very similar to the behavioral
gust front mask values, while the wind shear predicted values were slightly different
than the behavioral values.
Figure 7. Three
subjects’ individual reaction time performance results for the uniform fixed
pattern, wind shear and gust front conditions.
CONCLUSIONS
The luminance contrast
metric appears to do a fair job of predicting the results of this experiment
with the masking threshold estimated from the experiment. This metric would be
much more useful if the masking threshold were close to that found in other
studies, such as that of Scharff and Ahumada (2000). They used a value near 0.05 rather than 0.5. There are several reasons that the metric
might not be doing better here. The
metric is designed for predicting the effects of spatially homogeneous maskers,
it is not designed to predict the effects of variations in the mean luminance,
which are pronounced in the wind shear mask and which was better fit by an even
larger value for the masking threshold.
The metric as implemented also does not take into account the masking
effects of the text blocks themselves, which would be expected effectively to
raise the masking threshold.
ACKNOWLEDGEMENTS
A
special thanks to Dr. Jason McCarley for providing the software to run the
study and thanks to Kenneth Allendoerfer at the William J. Hughes Technical
Center and Dino Piccione at the Federal Aviation Administration for providing
the test stimuli and the invaluable support in accomplishing this study. Assistance in the data analysis and
calibration were provided by Jing Xing and Lori Shird, supported by NASA Ames
Research Center cooperative agreement NCC 2-1095 with the San Jose State
University Foundation.
The views expressed in this article are
those of the authors and do not reflect the official policy or position of the
Federal Aviation Administration, United States Department of Transportation,
nor the United States Government.
REFERENCES
Cole, R.E. and Wilson, F.W. (1994). The integrated terminal weather system terminal winds product, MIT Lincoln Laboratory Journal, 7(2), 475-502.
Scharff, L.V., Hill, A., Ahumada, A.J. (2000). Discriminability measures for predicting readability of text on textured backgrounds, Optics Express, 6(4), 81-90.
Wickens, C.D., Mavor, A.S. and McGee, J.P. (1997). Flight to the future: Human factors in air traffic control. National Academy Press, Washington, D.C.