DETECTION OF BLURRED CRACKS:
A STEP TOWARDS AN EMPIRICAL VISION
STANDARD
Bettina
L. Beard1, Karen M. Jones2, Cynthia Chacon2
and Albert J. Ahumada1, Jr.
1NASA Ames Research
Center, Moffett Field, CA 94035-1000,
2San Jose State
University Foundation, San Jose, CA
ABSTRACT
There
are no government mandated vision standards for aviation maintenance
inspectors. Empirically derived vision standards for other occupations cannot
be extended to this very different occupation. One important maintenance task
is the detection of metal fatigue cracks. To assess the effects of lowered
visual capacity on this visual detection task, we measured detection
performance by aircraft maintenance inspectors as a function of image blur. The
data are used to estimate the effect of blur-induced acuity decline on crack
detection probability, and provides empirical support for the construction of a
task-relevant visual acuity standard.
INTRODUCTION
It
is difficult, if not impossible, to eliminate human error in the process of inspection. Interventions must be developed to reduce
these errors and make the process more error-tolerant. Since visual inspection represents a large
part of aviation maintenance inspection, one mitigation strategy is to define
vision standards for this vision-intensive, safety-critical occupation. A
fine-tuned ability to localize, detect, discriminate, and identify job-relevant
stimuli can bring cost savings and safety benefits to industry.
In
2001, an FAA Advisory Circular (AC No: 65-31) recommended examination
guidelines for the vision of non-destructive inspection (NDI) personnel. It was
suggested that near and far vision in at least one eye must be 20/25 and 20/50,
respectively. Both near and far
requirements could be met with corrected or
uncorrected vision. This FAA
recommendation was based on acuity standards defined in other NDI/NDT
occupations.
No
current general standard exists in the aviation industry for the visual
qualifications of aircraft maintenance inspectors. Some maintenance facilities use the visual acuity and color
vision standards suggested in the FAA Advisory Circular, while other facilities
have defined their own vision requirements.
This illustrates the need for a uniform and universally accepted set of
vision standards that would apply to all aircraft non-destructive inspection
and testing (NDI/NDT) personnel.
There
are several broad steps that should be taken toward setting an objective and
empirically-based occupational vision requirement. The first step is a thorough vision task analysis. In the current
context, the FAA commissioned CAMI to perform this analysis focusing on the
role of visual processes. Next, to see
if a rigorously defined standard can be borrowed from a similar occupation, a
review of the literature should be undertaken.
Beard et al. (2002) compiled a
review of a text and WEB-based search for occupational vision requirements,
knowledge gained from site visits to major aircraft maintenance facilities,
relevant information from technical, mechanical, and inspection textbooks, the
FAA maintenance human factors web-site[1],
and the human vision literature. Beard et
al. (2002) found no studies that allow
generalization of standards to aircraft maintenance inspection. It is
unknown how similar tasks must be to validly borrow standards from another
occupation without being subject to compromise. What is needed is a rapid,
empirically-based methodology for defining occupational vision standards.
If
the standard cannot be legitimately borrowed from a previous standard, an
objective research methodology should be followed. In their review of the
vision standards literature, Beard et al. (2002) identified four occupations
that had empirically derived standards.
These empirical methodologies ranged from mathematically measuring the
size and working distance of the critical visual details (Sheedy, 1980) to
psychophysical measurements with blurring lenses placed in front of the eye on
a single task (Good & Augsburger, 1987; Good et al., 1996) or multiple
tasks (Padgett, 1989).
Here
we present a strategy for defining a visual acuity standard that permits
increased experimental control by blurring the image before presenting it to
the observer, within a computer program.
In this way what is done to the signal is exactly known.
The
primary objective of this research is to aid in the development of
recommendations for visual acuity requirements for aviation inspection
personnel. Specifically we determine that visual acuity deficits reduce
critical task performance and show in graphical form the relationship between
acuity decline and performance.
METHODS
Choice of a critical vision task for inspectors
The central question that must be addressed is
“At what level of visual deficit would a maintenance or inspection worker
become unable to safely and efficiently perform the critical visual tasks
required by the job?” Aircraft inspection is a complex process, requiring many
tasks, skills, and procedures. There
are multiple critical vision tasks that the workers are required to perform. One purpose of inspection is to detect
surface discontinuities such as cracks within the airframe and powerplant
regions of the aircraft. Cracks are
typically caused by two surfaces being overlaid at a boundary (Hellier,
2001). Since these cracks may be very
small and of low contrast, adequate visual acuity is likely to be involved in
their detection. After consultation
with domain experts, crack detection was chosen as the representative task in
order to ultimately set a visual acuity standard for aircraft maintenance
inspection.
Psychophysical Experiments
Observers
Two
female non-inspection personnel (age range from 23-30) and seven male
maintenance inspectors (age range from 35-58 years) participated in the
study. Maintenance inspectors were actively
employed and had from 10-18 years on the job.
All wore corrective lenses, though not always while inspecting. Near and far visual acuity, stereo vision,
and color vision tests revealed that all had at least 20/20 acuity, good color
vision, however one inspector lacked stereo vision.
Stimuli
Airframe and powerplant crack images were
obtained from various sources. Color images were converted to 8 bit
black-and-white images to delete any color cues. Before the experiment, “crack removed” stimuli were
generated. Using PhotoshopTM,
the crack was deleted from the image while maintaining the integrity of the
background image.
A difference image (the crack alone) was
generated by subtracting the no-crack image from the image with the crack. A “background-with-crack” image at contrast
level C was generated by multiplying the difference image by a C (<=1) and
adding it back to the background image.
The contrast in dB is 20 log10C. An image with a contrast of 0 dB has the original crack. An image with a crack contrast of 6 dB has
the difference image reduced by a factor of C = 0.5. This logarithmic scale keeps the variation in the results more
constant at different threshold levels.
Crack length estimation
To accurately determine the crack length and width,
estimates of the magnification in each photo had to be determined. Each photo included a circular label or
‘sticky’ whose diameter is a known 0.75 in.
To estimate the image magnification, PhotoshopTM was used to
identify the coordinates of six points along the perimeter of the sticky. These estimates of the perimeter were taken
by eye; therefore the error in these judgments was also determined. A computer program took these data and
computed a magnification value estimating the diameter of the sticky. When the sticky was on a flat surface, the
image is an ellipse and the estimates were very accurate. Some of the stickies were located on an edge
or curved surface. In these cases, coordinates were identified only on the flat
portion of the sticky and the ellipse estimated based on this flat
portion.
Figure 1: Crack length
and width estimates. Each image included a circular label or ‘sticky’ whose
diameter is a known 0.75 in. On the
left the sticky is located on a flat surface.
On the right the sticky is located on a curved surface. A magnification value estimating the
diameter of the sticky was computed from six points along the perimeter of the
sticky (shown in the figure).
At first, images were adjusted so that all the
stickies had a diameter of 0.75-in on the experimental display screen (the
actual size), but these images were so coarse because of display resolution
limitations that features of the fine cracks disappeared. Images were then adjusted to a screen sticky
size of 3-in, resulting in an average image width reduction from 1500 pixels to
800 pixels. Some of the images were
still larger than the screen resolution of 1024 by 768 and so were cropped to
990 x 660.
Apparatus
Photographs of large engine airframe cracks were
presented on a 1024x758-pixel display screen (SONY Trinitron). Viewing
was binocular with natural pupils. From observations of aircraft inspectors
performing primary inspections, Good (personal communication) found that the
majority of visual observations were done in the distance range from 34 to 40
cm. Because of screen resolution
limitations, images were magnified by 4 as discussed
above and so the experimental distance was comparably increased to 160 cm. From this distance each pixel subtended
0.31 arc min. The display background screen had a
mean luminance of approximately 40 cd/m2. Three lights illuminated a gray wall behind
the monitor. Another lamp illuminated
the ceiling behind the observer to achieve ambient lighting.
Photometric measurements of the SONY monitor revealed that screen
luminance values remained constant only after it was turned on for at least 45
minutes.
Figure 2: Example of an image with crack (upper left) and without crack
(upper right) images. The two lower
panels show these images after they have been blurred to simulate lower visual
acuity.
Simulating Visual Acuity Decline
Although the shape of the human blur function
differs between individuals and changes for different optical conditions, it
can be approximated by a Gaussian blur function. An observer with 20/20 visual acuity was assumed to have a
Gaussian blur spread[2]
of 2 arc min (Barten, 1999; Ahumada, 1996).
A person is said to have 20/40 visual acuity if they see at 20 ft what a
20/20 person sees at 40 feet. If we
assume that the 20/40 person has the same contrast sensitivity as the 20/20
person, then the blur for the 20/40 person must be twice the blur of the 20/20
person. Therefore, to simulate 20/40
visual acuity the combined blur of the image and the observer should be 4 arc
min. The combination rule for Gaussian
blur is the Pythagorean rule, so, for example, to obtain an acuity value of
20/40, the image blur spread was set to 3.46 since the Ö(3.46 2 + 2 2)
is 4. To obtain an acuity of 20/A where
A = the desired acuity level, then the blur in minutes = 2 Ö ((A/20)^2 – 1). Figure 2 presents example “crack removed”
and background-with-crack images with and without blur.
Procedures
Crack Contrast Detection
Thresholds
To increase the number of images tested and the
range of conditions, the two non-inspector observers collected data on a large
set of crack images at a greater number of blur levels, while the NDI/NDT
inspectors were run on subsets of crack images and blur levels.
Contrast
detection thresholds were obtained using a two interval forced choice staircase
method. The background airframe image remained on during the duration of the
block of trials. On a single trial,
observers were presented with the background alone in one 0.5 sec time interval
and the background with crack in another 0.5 sec time interval. The interval containing the crack was
randomized. The two time intervals were
demarcated with a simultaneous tone. Interval one contained one tone burst,
while interval two contained two tone bursts. Only one of the time intervals
contained the crack stimulus. The
observer’s task was to choose which interval contained the crack stimulus by
pressing one of two keys. The inter-stimulus interval was 0.5 sec. The sequence of each block of trials and the
crack with background image were randomly chosen.
A
different airframe image was presented in each block of trials, selected by a
random permutation of all of the images and blur levels to be presented in a
replication, for at least three replications.
To help the observer find the crack, in initial practice trials the
crack position was indicated to the observer by surrounding the crack with a
rectangle. After localizing the crack,
the observer could then practice the crack detection task without the surrounding
rectangle before continuing on to the experiment.
On
the first trial of a block of trials, the crack stimulus was presented above
threshold. Estimates of these supra-threshold contrast levels were determined
from model predictions (see Ahumada & Beard, 1998) and pilot data. The contrast was adjusted by a staircase
procedure. On each trial, if the
observer correctly responded as to which interval in which the crack was shown,
then the response was tallied as correct.
After three consecutive correct responses, the crack contrast was
decreased by a specified amount (step factor).
If the observer chose the interval that did not contain the crack
stimulus, then a brief feedback tone would sound, the response was tallied as
incorrect, and the crack contrast increased by a specified amount on the next
trial. To more rapidly converge to
threshold, initially the contrast step factor was 2 dB, but was reduced to 1 dB
after a change in the direction of the staircase (a reversal), and then reduced
to 0.5 dB after the second reversal.
After eight reversals in contrast and at least 30 trials, but no more
than 50 trials, the block of trials was terminated and the detection threshold
calculated by a probit analysis for that crack with background image.
The
two non-inspectors collected data on 10 images. The seven highly experienced aircraft maintenance inspectors
collected data on either a subset of these same 10 images or on 5 different
images. Observer CA collected data on
images at six levels of blur or acuity levels: 20/20, 20/25, 20/30, 20/35,
20/40, and 20/50. Observer KJ saw these acuities plus the acuity level
20/45. The 7 maintenance inspectors
collected data on 4 acuity levels: 20/20, 20/30, 20/40, and 20/50. To evaluate
the effect of viewing distance on the detection thresholds, one NASA observer
was run on a subset of her conditions at a farther viewing distance of 267 cm.
Contrast Sensitivity Functions
To estimate the observer’s internal blur and
screen resolution limitations, each observer’s contrast detection thresholds
were measured for a range of stimuli.
The Contrast Sensitivity Function (CSF) provides an estimate of visual
acuity because an individual’s resolving power is indicated by the intersection
of the curve on the abscissa of the graph. Horizontal and vertical contrast
thresholds were obtained to estimate orientation differences in the amount of
blur within the experimental display.
Much
like the experimental task, the observer had to decide in which of two 500 msec
intervals the stimulus was presented and respond accordingly (i.e., they
responded by pressing ‘1’ if they thought the stimulus was presented in
interval one, and ‘2’ if they thought the stimulus was presented in interval
two.) There was a 300 msec gap between
the presentations of the two stimulus images.
Instead of cracks, however, the target stimuli for this experiment were
a square, a line and a dipole. Observers completed this experiment while
sitting at a distance of 273 cm from the screen.
RESULTS
Probit analyses were done on each block of
trials to estimate the contrast threshold, the value at which the probability
of correctly identifying the interval was 75%.
The median of the scores replicating a particular condition was then
computed. In Figure 3, detection thresholds
are presented across blur or simulated acuity levels. Each symbol represents a different airframe image.
Figure
3: Contrast thresholds are presented across blur or simulated acuity
levels. Each symbol represents a
different airframe image. The results for observers CA and KJ are shown.
Figure
4 presents contrast thresholds for the different images as a function of blur
averaged over the four inspectors of group 1 and two of three inspectors in
group 2. The data of one inspector in
group 2 was not included because he was unable to see the cracks at the higher
blur levels. Each symbol represents a
different airframe crack. The effect of blur tends to be larger for cracks with
higher thresholds. The two images with
the highest thresholds could not be run at the higher blur levels
Figure 4: Contrast
thresholds for the different images as a function of blur averaged over
inspectors. Each symbol represents a
different airframe image.
Figure 5 shows the effects of blur on observers
averaged over images. Data for images
that were visible for all blur levels are shown. The effect of blur is quite
consistent over inspectors. The two non-inspectors (CA and KJ) showed lower
detection thresholds than did the experienced aircraft inspectors. A likely reason is that the non-inspectors
had participated in a study of practice effects on contrast thresholds using
similar images (Beard, et al., in preparation).
Figure 5: Effects of
blur for each observer averaged over images.
Observer initials are shown in the legend.
All data presented thus far were collected at a
distance of 160 cm. Because all
inspections are not done from one single distance, thresholds were measured
from a second distance of 266.8 cm. Thresholds were elevated at the further
distance, but show a similar increase in threshold with increases in blur.
Figure 6 shows the effect of increasing the
viewing distance. Thresholds for the
far distance are consistently higher than those for the nearer distance. If the
detection were simply a function of target contrast energy, the threshold would
be expected to increase by 20 log10(267/160) = 4.4 dB.
Attenuation of the high spatial frequency energy should cause an
additional increase in the threshold, which should be greater for the less blurred
stimuli and the higher threshold stimuli.
To foster translation
of these data into an occupational visual acuity standard, in Figure 7 we have
transformed the data from Figure 4 into Probability of Detection (PoD)
curves. The data were converted back to
probability of Yes/No detection after being normalized by setting the
probability of detection for 20/20 to 0.99 or 0.90. This calculation depends strongly on the assumed slope of the
psychometric function. Here we assume
the standard deviation of the cumulative Gaussian is 4 dB, but the actual value
could be anywhere from 1 dB to 6 dB.
Figure 7: Average data from Figure 4 and a
linear fit to that data (left). Probability of Detection curves based on the
linear fit (right). The probability of a correct response at 20/20 was
arbitrarily either set to 0.99 or 0.90.
DISCUSSION
Although good vision is a vital qualification
for aircraft maintenance inspectors, no general standards for visual acuity
currently exist for this occupation. Vision standards from other occupations
cannot be “borrowed” to set a standard for maintenance inspectors because the
visual demands between occupations are dissimilar and the majority of
occupational vision standards are not empirically based (Beard et al.,
2002).
One way to look at the effect of not having
20/20 vision is to say that an inspector with 20/40 vision sees at 20 feet what
the 20/20 inspector sees at 40 feet.
That is to say that the 20/40 inspector has to be twice as close as the
20/20 inspector to make the same discriminations. When the viewing distance is halved, the foveal search area is
reduced by a factor of 0.25, so it would take about 4 times as much time to
search the same area with the same discriminative ability if there is an acuity
deficit to 20/40.
In this project we measured detection
performance on a representative task performed by aircraft maintenance
inspectors as a function of image blur. These measurements allow predictions of
the amount the probability of detection could change as a function of
blur. As shown in Figure 7, cracks
whose detection was initially at 99% could be greatly reduced by blur
corresponding to only 20/30 if the inspection situation was kept constant in
all other respects.
The amount of visible contrast energy in the
crack correlated well with the contrast thresholds for the crack (r = -0.89). However, the effect of the blurring on the
thresholds was much greater than the loss in visible contrast energy. For the two images with the greatest loss
in visible contrast energy (4.7 dB) at the 20/40 blur level, the average
threshold loss was 10 dB. Although this
may be in part due to a lack of experience with these blurred images, it is
also possible that the blur causes more problems with crack detection than
predicted by contrast energy loss alone, such as affecting the extraction of
edges. The loss in visible contrast
energy can be thought of as a lower limit for the effect of blurring.
Blurring is only one possible cause of lowered
acuity. Another possible cause is
decreased overall contrast sensitivity.
In this case, the predicted effects are expected to follow more closely
the rule that cutting the viewing distance in half will compensate for a 6 dB
loss in sensitivity.
Methodological Limitations and Strengths
The experimental image generation procedure was
only an approximation of actual visual inspection. Inspectors were able to use
only one very relevant strategy (contrast detection) to look for the defect
embedded within a number of realistic aircraft locations. Although the cracks were positioned on
actual aircraft structures, inspectors could not use many of the common
strategies used in their work environment, such as tribal knowledge (knowing
where to look), moving closer, use of shadows (i.e., changing the angle of
light from their flashlight), touching the crack. But there is a trade-off between being able to use these
techniques and the time it takes to do a search.
Differences between the background conditions
indicate the effect of background variations on performance and will reduce the
importance of decision strategies on defect detection. This methodology permits manipulation of
defect absence, length, color, and other attributes. It is important to be capable of manipulating the absence of a
defect since uncertainty plays a large role in maintenance inspection (i.e.,
there is no prior knowledge that a defect will be present). In fact, it is only occasionally that a
defect is actually present.
Vision is a fundamental component of effective
aircraft inspection. All the same, so too are other cognitive
factors such as attention, memory, and experience. Inspectors are knowledgeable about individual components as well
as the overall aircraft being inspected, thus they possess the background to
properly locate, identify, and evaluate aircraft defects. Often NTSB accident
reports will point at visual deficits as contributors to accidents because a
crack went undetected, or a worker failed to detect fatigue damage. However, it may not be that vision led to
these overlooks. Other cognitive
factors may have played major roles in the lack of detection: job-related
stress, worker fatigue, multi-tasking, or memory effects of interruptions. The proposed research isolates vision
requirements on these duties. Because
the job entails much more than vision, these results may not relate to how well
the inspector will do on the job. Therefore, although vision is a critical
component in inspection, other factors weigh in heavily in the naturalistic
task.
Other requirements should address the effects of
other cognitive contributors. These
data can then be used by the FAA to write acceptable cognitive and perceptual
standards and procedures for inspectors including the type and frequency of
vision testing necessary to ensure the safe and effective performance of
current employees and job applicants who will perform a particular inspection
procedure.
Although psychophysical human-in-the-loop
experiments can provide accurate and objective data toward setting a standard,
it would be optimal to be able to predict performance using a computational
model. Ahumada, Beard, & Jones
(2005) show that a model of image discrimination does predict similar blur
effects as reported for model predictions of simulated crack stimuli (Beard et
al., 2003) but under-predicts the blur effects seen in psychophysical data
using these actual crack stimuli.
Guidance toward the setting of a standard
These measurements do not provide a standard,
but it converts the problem to specifying a desired physical limitation in
performance. The final step in the process
of defining a visual acuity standard lies in the hands of the FAA. Using the data in Figure 7, the FAA must
decide which stimulus characteristics and what margin of error (e.g., 1 error
in one million) will define where to draw the line for the standard.
Recruitment,
testing, selection, and training costs are high. The rejection of qualified persons imposes an unnecessary cost on
maintenance facilities. While the
failure of proper performance on visual tasks could be catastrophic, persons
with refractive errors such as correctable myopia who can perform the job
should be permitted to do so. Vision
requirements should be based on a demonstration that, for example, 20/25 near
or 20/50 distance visual acuity is actually needed to perform the essential
task. If the task is not generally
performed alone (i.e., there are several people in close proximity who provide
assistance) then these tasks should not be imposed with a vision requirement
for all the individuals. In addition,
vision requirements must be based on tasks that cannot be modified by current
available technology to assist the vision of the worker.
The
governing body, here the FAA, should clearly define the purpose of any vision
test and not provide medical examiners considerable latitude when conducting
visual acuity testing and evaluation.
An interesting case where this was not done, highlights the importance
of this recommendation. In a Safety Advisory entitled ‘Determination of Vision
Impairment among Locomotive Engineers” (SA-98-1) published by the Federal
Railroad Administration (FRA) and the Department of Transportation (DOT), a
lesson can be learned for the current purpose.
The FRA’s expectation was that the physicians who would be designated as
railroad medical examiners would be trained to competently administer color
vision examinations. Thus, they did not
anticipate that it would be necessary to specify for the medical examiners the
test procedure to be employed when testing for whether a person meets the
standards specified in this rule. That
assumption has been called into question under tragic circumstances. If the current rule had been implemented as
the FRA expected, the rule would have been adequate to prevent a major railway
accident involving the fatal collision between two New Jersey transit commuter
trains (NTSB/RAR-97/01). The NTSB
report found that the medical history of the suspect engineer showed that he
had been administered an acceptable test annually by the same contract physician
for over 10 years. In the tenth year, the test results showed a deterioration
of the engineer’s ability to distinguish among some colors. The engineer was then given a Dvorine
Nomenclature Test to further evaluate his color vision. Many color weak individuals can identify the
names of colors by their brightness instead of their hue. The examiner failed to administer the
accompanying Dvorine Second edition color vision test, which measures color
discrimination abilities and therefore the results of the first test suggested
that the engineer did not have a problem.
It was ruled likely that the accident was preventable if the physician
had used a sound approach to measure the person’s ability to distinguish
colors.
Self-monitoring
Aircraft maintenance
inspectors as a group take great pride in their ability to detect defects. In addition, they care deeply about the
safety implications of their job. Many environmental and developmental
variables can affect visual sensitivity.
Changes in vision are typically slow and subtle and therefore not easily
identified by the individual. Long work shifts or age-related accommodative
changes can lead to eye strain, headaches, excessive rubbing of the eyes,
esotropia or exotropia, and reduced efficiency on the job. Without an objective measuring tool, workers
will not detect gradual changes in their vision. If you don’t see something,
you don’t know that you can’t see it (self-awareness). Providing the workers
with a method to self-monitor their visual acuity would enhance occupational
safety and safety in the NAS.
The Use of Colored
Lenses
In
the workplace, some maintenance inspectors wear corrective lenses that have
been tinted with a color. Typically, this color is yellow. The media of the
human eye bends the light of the spectrum differentially, depending on the
wavelength. Yellow light is focused on the retina, whereas blue light is
typically several diopters out of focus.
This is referred to as lateral chromatic aberration. By filtering out blue light with the use of
yellow tinted lenses, images will be more in focus (Wolffsohn, et al.,
2000). One might suspect that yellow
tinted lenses are a benefit to the wearer.
This is not always the case (Tredici, 2005). Unless the colored lens is of unusually high quality, the amount
of light reaching the eye is reduced, even though the perceptual experience
while wearing yellow tinted lenses suggests that the environment appears
brighter (Kelly, 1990). Therefore, in
terms of visual acuity, there is a trade-off between the reduction in the
amount of blue light reaching the retina and the lower yellow light level
produced by the filtering lens. Workers
who feel they benefit from such lenses might consider using an LED flashlight
where the blue LED could be turned off.
There would then be no loss or scattering of light by lenses.
ACKNOWLEDGMENTS
Supported by FAA Agreement DTFA-2045 to B. L.
Beard.
We thank the inspectors who participated in the
study. They were wonderful
psychophysical observers who always showed great professionalism and devotion
toward this effort.
REFERENCES
Ahumada, Jr., A.J. (1996) Simplified vision
models for image quality assessment, in SID International Symposium Digest of
Technical Papers, ed J. Morreale, Vol. 27, pp. 397-400, SID, Santa Ana, CA.
Ahumada, A.J., Beard, B.L. (1997). Image
discrimination models predict detection in fixed but not random noise. Journal of the Optical Society of America A 14,
2471-2476.
Ahumada, A.J., Beard, B.L. (2005) Modeling the
detection of blurred visual targets in non-homogeneous backgrounds, Vision
Sciences Society Abstract 485, p. 134. http://vision.arc.nasa.gov/personnel/al/talks/05vss/05poster2.htm.
Ahumada, A.J., Beard, B.L., Jones, K.M. (1998) A
simple vision model for inhomogeneous image quality assessment, in SID Digest
of Technical Papers, ed. J. Morreale, vol. 29, Paper 40.1, Santa Ana, CA.
Barten P.G.J. (1999) Contrast sensitivity of the
human eye and its effects on image quality. Knegsel: HV Press.
Beard, B.L., Frank, T.A., Ahumada, A.J. (2003)
Using vision modeling to define occupational vision standards, Human Factors
and Ergonomics Society Annual Meeting (October 15, Denver, CO).
Beard, B.L., Hisle, W.A., & Ahumada, A.J.
(2002) Occupational vision standards: A review. Report for FAA AAR-100.
Good, G. W. & Augsburger, A. R. (1987).
Uncorrected visual acuity standards for police applicants. Journal of Police Science and Administration, 15(1), 18-23.
Good, G. W., Weaver, J. L., & Augsburger, A.
R. (1996). Determination and application of vision standards in industry. American Journal of Industrial Medicine,
30(5), 633-630.
Hellier, C. J. (2001) Handbook of Nondestructive Evaluation. New York: McGraw-Hill.
Kelly, S. A.
(1990) Effect of yellow-tinted lenses on brightness, Journal of the Optical Society of America A 7(10)
1905-1911.
Padgett, V. R. (1989). Summary of empirical
validation of firefighter vision requirements: four field experiments. Paper
presented at the International Personnel Management Association Assessment
Council 13th Annual Conference held in Orlando, FL. Retrieved July 11, 2002, from http://home.earthlink.net/~medtox/firevis.html
Sheedy, J. E. (1980). Police vision standards. Journal of Police Science and Administration,
8(3), 275-285.
Tredici, T. J. (2005) Aerospace
Ophthalmology, USAF Flight Surgeon's Guide: Chapt. 8, Retrieved May 13,
2005 from http://wwwsam.brooks.af.mil/af/files/fsguide/HTML/Chapter_08.html.
Wolffsohn, J. S., Cochrane, A. L, Khoo, H.,
Yoshimitsu, Y., Wu, S. (2000) Contrast is enhanced by yellow lenses because of
selective reduction of short-wavelength light. Optometry & Vision
Science 77(2), 73-81.