Metodología, delimitaciones y otras consideraciones

From here, this thesis will explore each task and relate them to theHVS, to provide examples of the various types ofvisual searchin regular use.

Equipment Fault Checking

A specific example of equipment fault checking performed by manual inspection is the inspection of aeroplane fuselages for cracks, corrosion, dents and loose rivets (Drury et al.1990; Gramopadhye et al.2002). The image presented to the airframe inspector is sparsely, yet uniformly populated withfeatures, permitting anexplicit searchfocused on specific locations within the image. The number ofdistractorson the airframe is zero, because the targets forvisual searchinclude any item presented on the image.

Diagnostic Medical Scans

The analysis of radiographs— as used in breast-cancer screening— is an extremely difficultvisual searchtask. 10–30% of images thatfeaturecancers are not reported (e.g. Mello-Thoms, Nodine et al.2002). Mello-Thoms, Dunn et al. (2002, p. 1007) found that locations upon the radiographs that were fixated upon for longer than usual durations were indicative of problem areas within the image that warranted further attention. The predominant type ofvisual searchwith radiographs isinefficient search(Mello-Thoms2003; Nodine et al.2001, p. 6). The eye movements of experts when observing radiographs is different to novices. Expert radiographers appear to use fewer fixationsand cover less of the image with their eyes in anexplicit searchfor abnormalities (e.g.

Drew et al.2013, p. 270).

The number ofdistractorsdifficult to determine. Separating potential items of interest within naturalistic stimuli— imagery that is derived from the real world and hence noisier than simulated representations of the same kind— such as a radiograph, is very difficult because the indistinct items to search amongst in thestimuliconstitute conjunctivefeatures. The complexity of the target featuressuppresses the usefulness ofefficient searchin favour of the slower and more inaccurate inefficient search(Boot et al.2009, p. 950).

Wilderness Search and Rescue

The combination of aninefficient searchwith fewdistractorsdescribes the task behind this thesis:

WiSAR. As previously explored in chapter1, aspotter(part of aSARteam) observes large quantities of sparely populated aerial photography in the search for a small inconspicuous target of a missing

individual or parts of a crashed aeroplane (e.g. Goodrich et al.2007). The uncertain description of the target increases the chances ofdistractorsbeing mistaken for the legitimate human in need of assistance.

As the UAV transverses the terrain, it frequently does so using an optimal spiral search pattern (e.g. Quigley et al.2005, p. 3030) at a specific flight speed and altitude (e.g. Doherty and Rudol2007, p. 5)— neither of which are conducive to assistingvisual search.

Despite the difficulty of identifying a human in distress from anUAV,spottersare still the best method (Croft et al.2007, p. 165) for this task, despite the limited time, resources and other difficulties associated with the task (Stager1974, p. 52).

While some expertly trainedspottersutilise the explicit form ofvisual attention(Stager and Angus1975), most are volunteers with minimal training. Therefore mostspotterssubconsciously adopt thesaccade and fixate strategyfor finding the target within the wilderness.

Security Baggage Inspection

The inspection of baggage as it passes by on a conveyer belt is an extremely difficultvisual search task (e.g. Nakayama and Martini2011, p. 1535). The very lowtarget prevalence— chance of atarget appearing — of extraordinary items within suitcases leads to high error rates (e.g. Wolfe et al.

2007).

Aninefficient searchof the x-ray display is conducted, with relatively short saccades due to the the uncertainty as to the properties of suspect baggage (e.g. Huestegge and Radach2012, p. 1024).

The variable appearance of x-rayed baggage (i.e.naturalistic stimuli) also causes large variations in thevisual searchmethod employed by the security staff (e.g. McCarley et al.2004, p. 305).

Control Panel Observation

The design of control panels— such as a vehicle dashboard or a aeroplane cockpit— is still in its infancy with respect tovisual search. Recommendations for alerting systems (e.g. fault alarms) include placing important warnings towards the centre of the observer’s visual field, with lower

priority alerts located further away (Phansalkar et al.2010, pp. 494–495). Phansalkar et al. (2010, p. 495) recommended that alerts share as fewfeaturesas possible with the remainder of the display

to increase saliency — i.e. to reduce the number of common featuresbetween the targetand distractors. Such distracting designs are intentional; relying upon preattentive visionto provide oculomotor capture(Prinzel and Risser2004, p. 22) to direct the observer towards the alert.

However, despite this reliance uponefficient searchto bring a flashing warning light amongst many dark and therefore inconspicuousdistractors, such techniques do not always bring the alert tovisual awareness. For example, in a report by Sarter et al. (2007, pp. 355–356), aeroplane pilots failed to notice their navigation computer changing mode during flight, despite the indicator being very close to the centre of the pilot’sfield of view, and despite many of the pilots actually fixating on the indicator lights!

Quality Control Inspection

In contrast to the “security baggage inspection” where the items to be observed are arranged erratically, “quality control inspection” is associated with a conveyer belt filled with a regular, organised array of items. Therefore inspectors perform a series of discretefixationsin a regular scan-path(i.e. a consistent series of sequential eye movements), such as inspecting each bottle neck for defects because faults regularly occur at these locations (e.g. Saito1972, p. 147).

This regularity in thestimulileads to aexplicit searchstrategy amongst manydistractors. The regular scan-paths used by inspectors were found to play a significant role in failing to detect faulty items on the conveyer belt (e.g. Megaw and Richardson1979, p. 152).

Photography Library Inspection

When browsing through a large photograph album in search of a particular photograph, the target photograph is recognised quickly as itpops-outof the image array intovisual awareness(e.g.

Christmann et al.2010; Cooper et al.2006).

Increasing theset size— the number of simultaneous images present on the display— does not appear to affect the ability for observer’s to identify a specific target photograph (e.g. Cooper et al.

2006; Spence and Witkowski2013). This absent effect ofset sizeon target recognition performance (e.g. reaction time) indicates that anefficient searchis being conducted (e.g. Carrasco et al.1995).