Vous êtes sur la page 1sur 7

CAN THE DIFFICULTY LEVEL REACHED IN COMPUTER-BASED TRAINING

PREDICT RESULTS IN X-RAY IMAGE INTERPRETATION TESTS?

Stefan Michel, Dr. Marcia Mendes Adrian Schwaninger, Prof. Dr.


University of Applied Sciences University of Applied Sciences University of Applied Sciences
Northwestern Switzerland (FHNW) Northwestern Switzerland (FHNW) Northwestern Switzerland (FHNW)
School of Applied Psychology (APS) School of Applied Psychology (APS) School of Applied Psychology (APS)
Institute Humans in Complex Institute Humans in Complex Institute Humans in Complex
Systems (MikS) Systems (MikS) Systems (MikS)
Riggenbachstrasse 16, 4600 Olten Riggenbachstrasse 16, 4600 Olten Riggenbachstrasse 16, 4600 Olten
Switzerland Switzerland Switzerland
stefan.michel@fhnw.ch marcia.mendes@fhnw.ch adrian.schwaninger@fhnw.ch

and and and

Center for Adaptive Security Center for Adaptive Security Center for Adaptive Security
Research and Applications (CASRA) Research and Applications (CASRA) Research and Applications (CASRA)
Thurgauerstrasse 39 Thurgauerstrasse 39 Thurgauerstrasse 39
8050 Zurich, Switzerland 8050 Zurich, Switzerland 8050 Zurich, Switzerland

Abstract – As scientific studies have shown, the recurrent certification of x-ray image interpretation competency
performance of airport security screening officers in x-ray image are derived. Furthermore, large increases in detection
interpretation depends critically on individual abilities and visual performance were found when comparing performance
knowledge acquired through class-room, computer-based obtained at the three measurement dates, supporting results of
(CBT) and on-the job training. The effectiveness of adaptive earlier studies.
CBT in increasing the efficiency in x-ray image interpretation of Altogether, this study provides further evidence that recurrent
screening officers could be verified in several studies. adaptive CBT is a powerful tool for increasing screening
Individually adaptive CBT systems, like X-Ray Tutor (XRT), officers' x-ray image interpretation competency.
are constructed in such a way that with increasing performance
of each individual the difficulty level will increase and become Index Terms — Airport security, x-ray screening, computer-
more challenging with respect to three imaged-based factors: based training, difficulty level, detection performance.
viewpoint (depending on rotation of threat items in a bag),
superposition by other items and bag complexity. The focus of
this study was to examine to what extend the achieved training I. INTRODUCTION
level in XRT positively correlates with performance in the X-Ray
Competency Assessment Test (X-Ray CAT) or, more The identification of prohibited items, when x-ray screening
specifically, whether screeners that have acquired higher passenger bags, is a challenging task. Within only a few
difficulty levels also achieved better test results. Knowledge seconds a human operator must decide whether a piece of
about an acquired difficulty level could then, subsequently, luggage can be regarded as harmless or whether a prohibited
provide an indication about the screeners’ x-ray image item is contained. Although airport security technology has
interpretation competency. Furthermore, analysis of these evolved remarkably over the past years, the actual identification
correlations would suggest which XRT difficulty level screening of prohibited objects still rests with human operators (i.e.
officers should have mastered in order to be acknowledged as screening officers). Consequently, the most advanced
proficient, and for them to be well prepared to pass initial and machines become vile if these people are not selected and
recurrent certification with the X-Ray CAT. The latter is of trained appropriately. According to object recognition theories
particular interest as the X-Ray CAT is being used for and visual cognition, object shapes not similar to the ones
certification at several European airports. A further goal of this stored in visual memory are difficult to recognize (e.g., [1]; [2];
study was to reproduce results of previous studies confirming [3]). [4] have shown that the detection of forbidden objects in x-
the effectiveness of adaptive CBT for increasing detection ray images of passenger bags depends on knowledge-based
performance of screeners. The study was carried out at one and image-based factors. Many prohibited items are rarely
international airport. 199 screening officers underwent training seen in everyday life and might look quite different from reality
with XRT for 20 months. Assessments were taken before the when displayed as an x-ray image (e.g., electric shock devices
start, after 13 months and finally after 20 months of training. or improvised explosive devices (IEDs)). Therefore, knowledge
Analyses showed high positive correlations between the about which items are prohibited and what they look like in x-
achieved XRT training levels and the individual's performance ray images is required. Class-room, computer-based (CBT) and
in the X-Ray CAT. Based on these results, recommendations on-the-job-training can help attaining such knowledge
on which XRT level should be reached before initial and (knowledge-based factors). Moreover, the detection of

1
prohibited items can be impaired by different image-based and after the full 20 months of training. Information about each
factors. [4] have identified three image-based factors which are individual’s achieved difficulty level after the 20 months of
relevant for x-ray screening: viewpoint (based on the viewpoint training was extracted from training data.
of threat items in a bag), superposition of a threat item and bag
complexity ([4]; [5]; [6]; [7]; [8]). Depending on perceptual A. X-Ray Tutor (XRT) training system
experience and the ability to mentally rotate objects, rotated
threat items are more difficult to recognize (effect of viewpoint). The X-Ray Tutor training system (XRT) was developed by
Likewise, position of a threat item in a bag, its superimposition the Visual Cognition Research Group (VICOREG) at the
by other objects (effect of superposition) and number and types University of Zurich as a scientifically based training program to
of objects in a bag (effects of bag complexity) also affect the enhance x-ray image interpretation competency of screening
difficulty of identifying threat items ([5]). As results of different officers ([2]). In the training program x-ray images of passenger
studies demonstrate, coping with image difficulty resulting from bags are presented to the screening officers, whereupon they
these factors can be improved through training. However, a have to visually inspect the images, search them for threat
person’s visual abilities greatly influence and alleviate this items and, subsequently, take a decision on whether a bag can
learning progress. Achieving a good x-ray image competency be regarded as harmless or not. XRT automatically combines
becomes much more difficult and challenging if a person does fictional threat items with x-ray images of harmless passenger
not have the right visual abilities. In airport security screening bags using an individually adaptive algorithm. In the standard
large differences between people with regard to aptitudes and edition most of the threat items can each be depicted from up to
abilities exist ([9]). For this reason it is recommended to apply 96 different views. Objects depicted from an unfamiliar
selection test as part of pre-employment assessments, so that viewpoint are more difficult to recognize, hence, a large and
right from the beginning those people best suited for the job are representative image library of prohibited items depicted from
chosen ([7]). many different viewpoints is necessary to provide a good basis
Regarding knowledge-based and image-based factors, for training x-ray image interpretation competency ([2]). In order
studies have shown that individually adaptive CBT systems, like to make sure the screening officers learn a large number of
X-Ray Tutor (XRT), are very effective and efficient tools for different prohibited objects, threat items from different threat
increasing x-ray image interpretation competency of screening categories are integrated in the XRT image library. The XRT
officers ([9]; [10]). XRT is a scientifically based training program image library was developed in close collaboration with experts
which’s effectiveness has been proven in several studies ([2]; of Zurich State Police, Airport Division as well as other
[11]; [12]; [13]). XRT is based on findings about how the human organizations and is continuously being extended. The XRT
brain processes visual information, taking the three relevant professional edition (XRT CBS PE), for example, contains four
image-based factors into account. One core advantage of XRT times as many threat items as the standard edition. In this
is that it is individually adaptive and level based, meaning study, screening officers have still been training with XRT CBS
training sessions are created on each trainee’s individual SE.
performance and learning progress. XRT is individually adaptive, meaning that it automatically
In this study, our main aim was to investigate to what extent a adapts to the performance of each individual in regard to coping
screening officer’s achieved difficulty level in XRT after 20 with image difficulty. The XRT training starts by showing threat
months of training positively correlates with his/her performance items depicted in easy (canonical) views, with low superposition
in the X-Ray Competency Assessment Test (X-Ray CAT). This and low bag complexities. The difficulty of each of these factors
study sought also to examine whether the achieved difficulty increases depending on an individual’s learning progress.
level can give an indication about an individual’s x-ray image These parameters are calculated by an individually adaptive
interpretation competency. The X-Ray CAT is an instrument algorithm using objective measures of view difficulty,
used to measure x-ray image interpretation competency, i.e. superposition and bag complexity ([3]; [12]). The individually
how well people can visually process x-ray images of adaptive training algorithm determines specifically for each
passenger bags in order to detect threat items ([14]). We screening officer which items and views are difficult to
theorize that screening officers that have acquired higher recognize and adapts the training to his/her requirements in
difficulty levels also achieved higher detection performance such a way that missed objects will be presented to him/her
scores. A further aim of this study was to replicate results of more often. In order to enter a higher difficulty level a screening
previous studies confirming the effectiveness of CBT for officer needs to have seen a threat item for at least three times
increasing x-ray image interpretation competency of screening in this level and detected it correctly 2/3 of the time. This
officers ([2]; [11]; [12]; [13]). specific item is then regarded as learnt.
The 12 difficulty levels of XRT are based on view difficulty,
II. METHOD AND PROCEDURE superposition, and bag complexity as described in Table 1.
During training with XRT images are displayed on the screen
199 cabin baggage screening officers of one selected for a maximum of 15 seconds (standard setting). In case a bag
international European airport underwent training with the XRT is judged as NOT OK (containing a prohibited item) the
standard edition for cabin baggage screening (XRT CBS SE) screener has to identify the threat item in the bag by clicking on
for a duration of 20 months. The screening officers were it. After each response feedback is provided immediately,
instructed to conduct at least one 20 minute training session informing the screener whether the response was correct or
per week. Measurements of detection performance were not. If the bag did contain a prohibited item, detailed information
collected conducting the X-Ray CAT for cabin baggage about this item as well as its location in the bag can be
screening at three different measurement dates; before start of displayed ([2]). The effectiveness of individually adaptive CBT
training as a baseline measurement, after 13 months of training compared to conventional (not adaptive) CBT has been

2
investigated in previous studies (e.g., [13]). to use a score that considers the hit rate1 as well as the false
alarm rate2. The hit rate on its own is not a valid measure,
TABLE I since a candidate can achieve a high hit rate by simply judging
DIFFICULTY LEVELS IN X-RAY TUTOR.
all bags as NOT OK. For our analyses we used A' ([16]), a
Level Viewpoint Super- Bag measure of sensitivity which is commonly used for a variety of
position Complexity tasks including screener certification and competency
1 Easy Low Low assessments. A' is a nonparametric measure, meaning that the
2 Difficult Low Low computation of A' requires no a priori assumption about the
underlying signal and noise distributions. A' is a kind of average
3 Easy High Low
between minimum and maximum performance and can be
4 Difficult High Low calculated by the following formula ([17]):
5 Easy Low Medium
6 Difficult Low Medium 1 1
A
2 4 1
7 Easy High Medium
8 Difficult High Medium H is the hit rate and F the false alarm rate. If performance is
9 Easy Low High below chance, i.e. when H<F, the equation must be modified
([18]):
10 Difficult Low High
11 Easy High High 1 1
A
12 Difficult High High 2 4 1

B. X-Ray Competency Assessment Test (X-Ray CAT) Even though the computation of A' requires no a priori
assumption about underlying distributions, it does not mean
that this measure is an accurate reflection of its theoretical
The X-Ray CAT for cabin baggage screening (X-Ray CAT
origin (i.e., that A' reflects the area under a reasonable ROC
CBS) is a reliable and valid instrument to measure x-ray image
curve) or that it is a distribution-free measure and fully
interpretation competency of airport security screening officers.
independent of response bias (see [19]). However, due to its
It has been developed according to the principles and
easy computation and interpretation, A' is often applied as a
requirements specified in [15], and it is applied for screener
measure for detection performance in research and application.
certification at several European airports. It consists of 256
In this paper, actual performance values are not reported due
trials based on 128 different color x-ray images of passenger
to security reasons. For all relevant analysis effect sizes are
bags. All bag images are once displayed without (non threat
reported and interpreted based on Cohen ([20]).
image) and once containing a prohibited object (threat image).
The prohibited objects are grouped into four different categories
as defined in ECAC DOC 30: guns, IEDs, knives and other III. RESULTS AND DISCUSSION
prohibited items. Every category contains 16 threat objects,
subdivided into 8 pairs. Each pair is made up of two items
In this study, the relationship between x-ray image
visually similar in shape. One of the prohibited items is included
interpretation competency measured with the X-Ray CAT CBS
in the XRT image library (Set A), whereas the other item has
and the achieved difficulty level in XRT CBS SE after 20
not been seen before (Set B). This setup allows testing for
months of training was analyzed using data of 199 cabin
transfer effects of visual knowledge, i.e. whether training with
baggage screening officers. Moreover, a recommendation was
XRT also improves the detection of items that are visually
derived to determine which difficulty levels should have been
similar in shape but not contained in the XRT library and
mastered in order for screening officers to be well prepared to
therefore not known from training.
pass initial and recurrent certification. In addition to that, test
Furthermore, all items are once depicted in easy and once in
results taken at three different points in time were compared to
difficult view. Easy view refers to the canonical view, whereas in
investigate the effectiveness of XRT.
difficult views threat objects are shown in an 85° horizontal or
vertical rotation. Sets A and B are equalized with regard to the
A. Relationship between level and x-ray image interpretation
rotation of items.
competency
The X-Ray CAT takes about 30-40 minutes to complete.
While images are displayed for a maximum of 15 seconds
A high correlation between the achieved training level in XRT
screening officers have to take a decision whether a bag is OK
CBS SE and detection performance (A') after the 20 months of
(containing no prohibited item) or NOT OK (containing a
training was found with r = .64, p < .01. Figure 1 shows the
prohibited item). For more information on the X-Ray CAT see
[14].
1 The hit rate counts how often a person has correctly judged
C. A' as measure for detection performance a bag as NOT OK in proportion to all bags containing threat
items.
For a valid measure of detection performance it is important 2 The false alarm rate counts how often a person has
wrongly judged a bag as NOT OK in proportion to all bags
containing no threat item.

3
increase of detection performance as a logarithmic function of acquired a reliable x-ray image interpretation competency and
the XRT level. Lowest test results were performed by those be well prepared to pass initial and recurrent certification with
screeners in very low levels whereas highest detection the X-Ray CAT. The duration of how long it takes to reach a
performance was achieved by the screening officers having certain training level has been analyzed in a previous study
acquired level 12 (the highest level in XRT CBS SE). A general ([21]). According to their results, if screeners use the training
improvement in detection performance can be observed for system 20 minutes per week then one year of training is
each rising level. Yet, it must be considered that quite required to reach level 6.
remarkable dispersions exist between the persons with the
highest and lowest detection performance within each level. B. Effect of training

1 Similar to previous studies large increases in detection


performance were found for the three test dates. Figure 2
Detection Performance (A')

shows means and standard deviations3 broken up by category


0.9
and test date.

0.8 1st data collection 2nd data collection 3rd data collection

Detection Performance (A')


0.7

0.6
y = 0.079ln(x) + 0.687
R² = 0.471
0.5
1 2 3 4 5 6 7 8 9 10 11 12

XRT Difficulty Level

Fig. 1: Correlation between detection performance A’ and XRT


CBS SE difficulty level. Guns IEDs Knives Other

For initial and recurrent certification we recommend a pass


Fig. 2: Mean detection performance A‘ and standard deviations
mark of A' >= .80 when applying the X-Ray CAT. Of the 199
broken up by threat category and test date. Note: For security
screening officers participating in this study, 37 (18.6%) did not
reasons A’ scores are not indicated in the figure.
reach this pass mark. As can be seen in the graph, most of
these screeners were still training in lower XRT levels and had
Substantial increases in detection performance could be
not yet reached level 6.
observed for all categories, especially for IEDs.
According to [3] prohibited objects in x-ray images can be
A repeated measures analysis of variance (ANOVA) using A’
recognized best when they have already been seen from
scores was conducted with test date (first, second, third) and
different viewpoints, superimposed by other objects and
category (guns, IEDs, knives, other) as within-participants
presented in more or less complex bags. As described above,
factors. There were large main effects of test date F(1.66,
the 12 difficulty levels of XRT CBS are build up in such a way 2
329.46) = 196.25, p < .001, η = .50, and category F(2.63,
that in level 1 only bags with items depicted in easy viewpoints, 2
521.61) = 235.07, p < .001, η = .54, and a large two-way
with low superposition and low bag complexity are displayed,
interaction of test date and category F(4.82, 953.31) = 35.24, p
whereas with rising difficulty level the image-based factors, 2
< .001, η = .15. These results are consistent with earlier
become more challenging (see Table 1). Once level 6 has been
studies ([2]; [11]; [12]; [13]) showing that XRT is a very effective
reached, a screener has seen threat items from easy as well as
tool to increase x-ray image interpretation competency.
difficult viewpoints, with both low and high superposition.
To examine the effect of viewpoint on detection performance,
Additionally, the threat items have been presented in bags of
another ANOVA for repeated measures was conducted. A’
low and medium complexity. Screeners have therefore been
scores for both types of views (easy and difficult) with the
exposed to a good variation of image-based factors. Taking
within-participant factors test date (first, second, third), view
these considerations into account, it can be expected that once
(easy vs. difficult), and category (guns, IEDs, knives, other)
having reached level 6 in XRT CBS SE screeners should be
were used. Large main effects of test date F(1.69, 334.30) =
able to reliably recognize threat objects in x-ray images. 2
213.10, p < .001, η = .52, view (easy vs. difficult) F(1, 198) =
The results of this study support this theory. Almost all 2
887.70, p < .001, η = .82, and category F(2.67, 529.23) =
screeners who had reached XRT level 6 achieved good test 2
232.60, p < .001, η = .54 were revealed. Interactions were
results in the X-Ray CAT, high enough to pass certification
significant and there were large main effects for the two-way
when A’ >= .80 is used as a pass mark. Of all screening officers
who had acquired XRT level 6, only 6 persons (4.3%) had an A’
score < .80. 3 The standard deviation represents the range of dispersion
All in all, it can therefore be expected that once having around the mean data and indicates the range of individual
reached level 6 in XRT CBS SE screening officers should have differences between the tested airport security screeners.

4
interactions between view and category F(3, 5 594) = 129.61, p < with XRT not only had an effect on the t detection of items that
2
.001, η = .40 and date and category F(4.87, 9 963.93) = 33.01, p were included in the XRT training librrary but also led to a better
2
< .001, η = .14. For the two-way interaction of date and view detection of untrained and unfamiliar objects. Apparently, visual
2
F(1.72, 341.20) = 23.12, p < .001, η = .11 a m medium effect was knowledge acquired through XRT tra aining could be transferred
revealed. The three-way interaction of testt date, view and to new objects that were similar in sha
ape.
category was significant and had a small e effect size F(5.33,
2
1055.70) = 9.31, p < .001, η = .05. Figure e 3 visualizes the
effect of test date for each category in both easy and difficult
views. It demonstrates that for every catego ory threat objects
were easier to detect in easy (or canonical) views than when
depicted from a difficult view. This applies parrticularly to knives.
However, as results confirm, recurrent CBT T with XRT could
enhance detection performance of screenerrs very effectively
even when threat objects were depicte ed from unusual
viewpoints.

Fig. 4: Detection performance A‘ and standard deviations


broken up by threat category, test datte and set (A vs. B). Note:
For security reasons A’ scores are not
n indicated in the figure.

IV. SUMMARY AND CONCLUSIONS


C

The main aim of this study wa as to investigate whether


performance in the X-Ray CAT can be related to the difficulty
level that screening officers have ac chieved in XRT. With this
information a recommended minim mum difficulty level was
Fig. 3: Detection performance A’ and stan ndard deviations derived that indicates whether sc creening officers can be
broken up by threat category, test date and d view (easy vs. acknowledged as proficient and we ell prepared to pass initial
difficult). Note: For security reasons A’ scoress are not indicated and recurrent certification. The analyssis of the results showed a
in the figure. high correlation between the XRT X training level and
performance in the X-Ray CAT afte er 20 months of recurrent
In order to test whether the visual knowledgge gained through CBT. In other words, screening officers
o who had reached
training with XRT could also be transferred tto items that were higher XRT difficulty levels also achie
eved better X-Ray CAT test
not shown during training but similar in shape, detection results. As explained earlier, we recoommend a pass mark of A’
performance for sets A and B was compared d. Figure 4 shows >= .80 when using the X-Ray CA AT for initial or recurrent
the detection performance for each threat category at all three certification. The results from our stu
udy suggest that screening
measurement times broken up by test set. A clear increase in officers who have reached XRT le evel 6 are quite likely to
detection performance for every category in b both sets over the achieve this pass mark. Furthermore, once having acquired
three measurements could be observed d. The repeated level 6 in XRT a screening officer has been exposed to a large
measures ANOVA with the within-participantt factors test date variety of images of passenger bags s and a broad variation of
(first, second, third), set (A vs. B) and categ gory (guns, IEDs, image-based factors. According to [3] objects can be
knives, other) revealed large main effects forr test date F(1.67, recognized best when they have been b seen from different
2
330.70) = 203.07, p < .001, η = .51, set (A vvs. B) F(1, 198) = viewpoints. Thus, taking the setup of XRT into account,
2
57.05, p < .001, η = .22, and category F F(2.66, 525.94) = screeners who have achieved XRT T level 6 should be well
2
239.49, p < .001, η = .55. Interactions werre significant with trained and able to reliably recognize threat objects in x-ray
large main effects for the two-way interactio ons test date and images. Objects have then been seen s in easy as well as
2
category F(4.84, 957.63) = 37.34, p < .001, η = .16, and set difficult views, with low and high sup perposition and in bags of
2
and category F(2.76, 545.67) = 41.17, p < .001, η = .17. A low and medium complexities.
action of test date,
small significant effect for the three-way intera In sum, the results of these ana alyses suggest that those
2
set and category F(5.4, 1069) = 3.45, p < ..01, η = .02 was screening officers having acquired le evel 6 in XRT CBS SE can
found. Taking a closer look at Figure 4 it can n be observed that be regarded as proficient with resp pect to their x-ray image
detection performance for sets A and B were e very similar, with interpretation competency. Moreove er, they should be well
the exception of the category “others”. The revealed large main prepared to pass initial and recurren nt certification with the X-
effect for set therefore must be explained by the category Ray CAT when the pass mark is set at A’ >= .80. In a previous
“others”. Altogether, the results clearly demonstrate that training study [21] have shown that screenerrs reach XRT level 6 after

5
one year of training if the training system is being used as [2] A. Schwaninger, “Computer-based training: A powerful
recommended (i.e. 20 minutes per week). tool to the enhancement of human factors,” Aviation
In addition, we have compared performance at three Security International, vol FEB/2004, pp 31-36.
measurement dates in order to verify the effectiveness of
individually adaptive CBT with XRT. For every threat category [3] A. Schwaninger, “Objekterkennung und Signaldetektion:
large increases in detection performance were found after 20 Anwendung in der Praxis,“ In: B. Kersten & M.T. Groner
months of training. A particular improvement could be observed (Eds.), Praxisfelder der Wahrnehmungspsychologie,
for IEDs, indicating that a reliable detection of objects such as Bern: Huber, 2004, pp 106-130.
IEDs is not difficult as such, but largely dependent on training.
Furthermore, analyses revealed large effects of viewpoint, [4] A. Schwaninger, D. Hardmeier, and F. Hofer, “Aviation
showing that items depicted from unusual (rotated) viewpoints security screeners visual abilities and visual knowledge
are more difficult to detect than items depicted from an upright measurement,” IEEE Aerospace and Electronic Systems,
(canonical) viewpoint. Large improvements in detection vol 20(6), pp 29-35, 2005.
performance were found for every threat category in both
views, which provides further evidence for the effectiveness of [5] A. Schwaninger, “Evaluation and selection of airport
XRT for x-ray image interpretation training. Similar effects were security screeners,” AIRPORT, vol 02/2003, pp 14-15.
found when comparing detection performance for items shown
in training (set A) vs. items that had not been seen before, but [6] D. Hardmeier, F. Hofer, and A. Schwaninger, „The x-ray
were visually similar in shape (set B). As Figure 4 object recognition test (x-ray ORT) - a reliable and valid
demonstrates, detection performance for both sets A and B instrument for measuring visual abilities needed in x-ray
was very similar, with the exception of the category “others”. screening” in IEEE ICCST Proceedings, vol 39, pp 189-
Remarkable improvements in detection performance 192.
throughout the three measurement dates could be observed for
every category. Thus, visual knowledge gained through training [7] D. Hardmeier, F. Hofer, and A. Schwaninger, “Increased
with XRT could also be transferred to other objects. detection performance in airport security screening using
All in all, findings of this study about the effectiveness of the x-ray ort as pre-employment assessment tool,” in
nd
recurrent adaptive CBT are fully consistent with results of Proceedings of the 2 International Conference on
previous studies ([11]; [12]; [13]). Nonetheless, increasing x-ray Research in Air Transportation, ICRAT 2006, Belgrade,
image interpretation competency of screening officers with CBT Serbia and Montenegro, June 24-28, 2006, pp 393-397.
does not mean that operational performance at airport security
checkpoints will be automatically increased as well. Adaptive [8] F. Hofer, D. Hardmeier, and A. Schwaninger, “Increasing
CBT is an indispensible prerequisite for good performance at airport security using the X-Ray ORT as effective pre-
th
security checkpoints, since it allows exposing screeners to employment assessment tool,” in Proceedings of the 4
objects they usually do not encounter in real life; however, it is International Aviation Security Technology Symposium,
not the only important performance factor. Additionally to CBT Washington, D.C., USA, November 27 - December 1,
class-room and on-the-job-training are necessary. Screening 2006, pp 303-308.
officers require theoretical knowledge and need to be trained on
how to behave adequately or how to deal with special situations [9] D. Hardmeier, F. Hofer and A. Schwaninger, “The role of
at the checkpoint. As mentioned earlier, large differences recurrent CBT for increasing aviation security screeners'
between people with regard to aptitudes and abilities in security visual knowledge and abilities needed in x-ray screening,”
th
screening exist. The usage of selection tests as part of pre- in Proceedings of the 4 International Aviation Security
employment assessment procedures can therefore very Technology Symposium, Washington, D.C., USA,
effectively help to improve performance in security screening November 27 - December 1, 2006, pp 338-342.
([9]). Moreover, people could easily get distracted or
unmotivated during work, hence, threat image projection (TIP) [10] A. Schwaninger, and F. Hofer, “Evaluation of CBT for
is a good technology to apply to increase and maintain increasing threat detection performance in X-ray
motivation and alertness of screeners. Also, as people often fail screening,” In: K. Morgan and M.J. Spector, The Internet
to react appropriately if something unexpected happens, Society 2004, Advances in Learning, Commerce and
frequent practical tests with real threat objects should be Security. Wessex: WIT Press, 2004, pp 147-156.
conducted regularly ([22]). Apparently, for the achievement of a
good operational performance at security checkpoints several [11] S. Michel, J.C. de Ruiter, M. Hogervorst, S.M. Koller, R.
factors play an important role and need to be taken into account Moerland, and A. Schwaninger, "Computer-based training
([22]). increases efficiency in x-ray image interpretation by
aviation security screeners,” in Proceedings of the 41st
V. REFERENCES Carnahan Conference on Security Technology, Ottawa,
October 8-11, 2007, pp 201-206.
[1] M. Graf, A. Schwaninger, C. Wallraven, and H.H. Bülthoff,
“Psychophysical results from experiments on recognition [12] A. Bolfing, T. Halbherr, and A. Schwaninger „How image
and categorization,” Information Society Technologies based factors and human factors contribute to threat
(IST) Programme, Cognitive Vision Systems - Cog Vis detection performance in x-ray aviation security
(IST-2000-29375), 2002. screening,” HCI and Usability for Education and Work,
Lecture Notes in Computer Science, vol 5298, pp 419-

6
438. (www.casra.ch) in Zurich. He manages different national and
international projects in aviation security which provide clients
[13] S.M. Koller, D. Hardmeier, S. Michel, and A. with cost effective methods to achieve security performance
Schwaninger, “Investigating training, transfer, and improvements and operational effectiveness. Stefan has a
viewpoint effects resulting from recurrent CBT of x-ray strong background in psychology with experience in aviation
image interpretation,” Journal of Transportation Security, security enhancing the interaction between people,
vol 1(2), pp 81-106, 2008. processes, and technology. Previously, he worked at the
University of Zurich (Switzerland) and the Max Planck
[14] S.M. Koller, and A. Schwaninger, “Assessing X-ray image Institute for Biological Cybernetics in Tübingen (Germany) as
interpretation competency of airport security screeners,” in a research scientist and received the doctoral degree in
nd
Proceedings of the 2 International Conference on 2008.
Research in Air Transportation, ICRAT 2006 Belgrade,
Serbia and Montenegro, June 24-28, 2006, pp 399-402. Marcia Mendes works as a research scientist at the
University of Applied Sciences Northwestern Switzerland
[15] A. Schwaninger, A.P. Brigdes, C. Drury, F. Durinckx, P. (FHNW) and the Center for Adaptive Security Research and
Durrant, T. Hodge, F. Hofer, R. Jongejan, R.L. Maguire, Applications (www.casra.ch) in Zurich. She is a member of
A. McClumpha, E. Neiderman, C. Steinmann, W. Wüest, the security research team led by Prof. Dr. Adrian
Principles and Requirements for Assessing X-Ray Image Schwaninger and involved in several national and
Interpretation Competency of Aviation Security Screeners, international projects in aviation security. Marcia finished her
White Paper, International Aviation Security Human studies in psychology (work and organizational psychology,
Factors Technical Advisory Group (InterTAG), health psychology, clinical psychology and public health) at
Competency Assessment Working Group (CAWG), 2005. the University of Bremen (Germany) in 2009. Currently, she
is writing her doctoral thesis in the field of aviaton security.
[16] I. Pollack, and D.A. Norman, “A non-parametric analysis
of recognition experiments,” Psychonomic Science, vol Prof. Dr. Adrian Schwaninger lectures at the University of
75, pp 125-126, 1964. Zurich since 1999 and at the University of Applied Sciences
Northwestern Switzerland since 2008. He is the head of the
[17] J.B. Grier, “Nonparametric indexes for sensitivity and bias: Center for Adaptive Security Research and Applications
Computing formulas,” Psychological Bulletin, vol 75, pp (www.casra.ch) in Zurich and the head of the Institute
424-429, 1971. Humans in complex Systems (MikS) at the School of Applied
Psychology, University of Applied Sciences Northwestern
[18] D. Aaronson, and B. Watts, “Extension of Grier’s Switzerland (www.fhnw.ch/miks). His areas of expertise are
computational formulas for A’ and B’’ to below-chance aviation security, human factors, scientifically based software
performance,” Psychological Bulletin, vol 102, pp 439- development, applied cognitive psychology, and human-
442, 1987. machine interaction. Adrian is a member of the ECAC
Training Task Force, the ECAC Technical Task Force, the
[19] R.E. Pastore, E.J. Crawley, M.S. Berens, and M.A. Skelly, ICAO Working Group on Training, and he leads the ECAC
“‘Nonparametric’ A’ and other modern misconceptions Technical Task Force TIP Study Group. Adrian is recognized
about signal detection theory,” Psychonomic Bulletin & as a leading authority on aviation security. He has more than
Review, vol 10, pp 556-569, 2003. 70 publications and more than 150 invited presentations. In
1999 he has received the Young Researcher Award in
[20] J. Cohen, “Statistical power analysis for the behavioral Psychology. In 2003 he has received the ASI International
sciences,” New York: Erlbaum, Hillsdale, 1988. Award of Excellence in Aviation Security: Enhancement of
Human Factors.
[21] S. Michel, S.M. Koller, and A. Schwaninger, “Relationship
between level of detection performance and amount of
recurrent computer-based training,” in Proceedings of the
nd
42 Carnahan Conference on Security Technology,
Prague, October 13-16, 2008, pp 299-304.

[22] A. Schwaninger, “Airport security and human factors:


From the weakest to the strongest link in airport security
screening,” Proceedings of the 4th International Aviation
Security Technology Symposium, Washington, D.C.,
USA, November 27 – December 1, 2006, pp 265-270.

VI. VITA

Dr. Stefan Michel is currently working as a scientific


researcher and project manager at the University of Applied
Sciences Northwestern Switzerland (FHNW) and the Center
for Adaptive Security Research and Applications

Vous aimerez peut-être aussi