Keywords

1 Introduction

In a modern warfare as well as in reconnaissance operations it is on one hand highly important to hide and protect own troops but on the other hand find and target the enemy. The protection of armament can be achieved for example by using camouflage and/or decoys. Most visual intelligence information gathered today comes from aerial photographs taken by various means of ISR (e.g., spy satellites, reconnaissance aircraft, UAVs) and presented in different displays. However, photos of dummy and/or camouflage tanks, planes and guns can deceive even trained analysts. In order to be able to hide and seek as well as build good decoys it is important to study the processes behind observing, detecting and identifying military objects. However, even though there is some previous research on camouflage (see for example King et al. 1984), there is scarcity of publically available studies especially on camouflage and decoys in military settings.

A military decoy is a low-cost fake military equipment device intended to represent a real item of military equipment to fool enemy forces into attacking them and so protect the real items of equipment by diverting fire away from them. A fake item is usually intended to be visible to enemy. In contrast, military camouflage is the use of camouflage to protect personnel and equipment from observation by enemy forces. In connection with tanks, camouflage differs from personal camouflage in that as the primary threat is aerial reconnaissance the goal is to disrupt the characteristic shape of the vehicle, to reduce shine, and to make the vehicle difficult to identify even if it is spotted. In practice, the main difference between real and fake and without and with camouflage objects is the lack in the details and roughness of shapes and contours visually available for the observer. Therefore, in general we expected that fake and camouflage images of tanks should be harder to detect than images of real and no-camouflage tanks.

On the basis of aforementioned arguments we expected that the sensitivity (i.e., signal-noise ratio) would be better in connection with real and non-camouflage images as compared to fake and camouflage images (Hypothesis 1A). We also expected that the identification speed would be lower and error rate higher for the fake and camouflage as compared to real and non-camouflage images (Hypothesis 1B). In connection with decoys, we expected that because decoys lack the visual details as compared to real objects, fake tanks would involve more attention to the image area and less attention switching between image and background than real tanks (Hypothesis 2A). We also expected that because camouflage disrupts the characteristic shape and contour of the vehicle as compared to the background, camouflage tanks would involve more attention to the outside image area and attention switching between image and background as compared to non-camouflage tanks (Hypothesis 2B).

We investigated the abovementioned hypothesis in two experiments: in the first experiment we examined the categorizations speed and accuracy of real and fake tanks with and without camouflage; in the second experiments we examined the eye-tracking activity associated to the evaluation of the images.

2 Methods

2.1 Participants and Materials

A total of 28 subjects participated to the study. They were 15 men and 13 women aged from 18 to 56 (M = 38). Most of them were associated with Finnish Defense Forces (as officers, civilian staff or conscripts). Eight people took part to experiment 1 (categorization speed) and 20 people (divided in two groups of 10 people) in experiment 2 (group 1 for images of real and fake tanks without camouflage and group 2 for images of real and fake tanks with camouflage).

On the basis of expert evaluations, we chose 9 pictures of real tanks and 9 pictures of fake tanks publically available in internet and standardized them to about a size of 600 × 400 pixels. The camouflage versions of the tanks were produced by 40% blending by a camouflage image with the tank images using Photoshop image processing software. The real (top-left) and fake (top-right) and their camouflage versions (bottom left and right) are illustrated in the Fig. 1.

Fig. 1.
figure 1

Images of the real and fake tanks without (top) and with (bottom) camouflage.

Full size image

2.2 Measures

Demographics (age, gender, etc.) and was assessed with a self-report questionnaire.

Go/No-go test (see e.g., Fillmore et al. 2006) in the Inquisit stimuli presentation software (millisecond.com) was used to assess the identification speed and accuracy of images of tanks in experiment 1. A sensitivity index (d prime), identification speed and error rates were calculated. The d prime sensitivity index was calculated as z-score of the overall successful hit rate minus z-score of the overall false alarm rate in the GO/NO test.

In the experiment 2, eye movements were tracked using a head-mounted eye-tracker “Dikablis” (Ergoneers GmbH, Manching, Germany) with a sampling rate of 60 Hz, scene camera field of view of 120°, using four point calibration, contrast pupil detection and D-Lab 3.5 recording software. Areas of interests (AOIs) were drawn with D-lab AOI tool for tank image and background areas and used to examine the visual activity with-in and between image and background by calculating the attention ratio and transition times between image area and background.

2.3 Procedure

Before experiment, the participants were properly informed and instructed in the course of experiment.

In experiment 1 (detection speed), the participant was told to respond as fast as possible to the target type of stimuli by pressing a space button in computer. Stimuli consisted of four blocks of trials. In first block the task was to response as fast as possible for the real tanks (but not for fake tanks); In the second block the task was to response as fast as possible for the fake tanks (but not for real tanks); Block 3 was same as block 1 and block 4 the same as block 2 but with camouflaged images. The order of the blocks was balanced using Lating Square. Each block consisted of a practice session followed by a trial with 50 stimuli (40 go stimuli and 10 no-go stimuli). The stimuli were randomly selected from the set of 18 images without and 18 images with camouflage.

In the experiment 2 (eye-tracking), participants were told to look at the images for six seconds as they were presented the center of the computer screen and then choose a between two radio buttons (real, fake) whether the image that was presented was a fake or a real tank.

After experiment the participants were debriefed and thanked. The experiments took place in a quiet office room and took from 10 min (experiment 1) to 20 min (experiment 2).

2.4 Data Processing and Analysis

Data for the first experiment (identification speed) was analyzed by the Linear Mixed Model in SPSS, with tank type (real, fake) and camouflage (without, with) as fixed factors and sensitivity index, identification speed and error rate, each in turn, as a dependent variable.

The eye-tracking data was analyzed by the General Linear Model (GLM) Repeated Measures procedure in SPSS, with image type (real, fake) and camouflage (without, with) as with-in subjects factors and continuous independent variables (i.e., AOI attention ratio and percentage of transition times), each in turn, as a covariate.

3 Results

As expected and illustrated in summary Table 1 below, the results showed that (1) the sensitivity was higher for real than fake tanks (for camouflage there was no difference), (2) the identification speed was lower for fake and camouflage images as compared to real and non-camouflage images, and (3) the error rate for fake images was higher than for real images (for camouflage vs. non-camouflage there was no difference). Therefore the results supported our Hypothesis 1A (partly) and 1B.

Table 1. Summary of the results
Full size table

As also expected and illustrated in the Table 1, awe found that fake tanks involved more attention to the image area and less attention switching between image and background than real tanks and that camouflage tanks involved more attention to the outside image area and attention switching between image and background as compared to non-camouflage tanks. Therefore our Hypothesis 2A and 2B were supported.

4 Discussion and Conclusions

In the present study we examined the perception, detection and identification of real and decoy tanks without and with camouflage. We found, among other things, that fake and camouflage images of tanks as compared to real and non-camouflage images decreased identification speed and that camouflage images elicited more attention shifting between image and background as compared to non-camouflage images. We argue that the results support that in general fake and camouflage images of tanks are harder to detect than images of real and no-camouflage tanks because they lack in the details and roughness of shapes and contours visually available for the observer. A follow-up study is now being conducted to look more thoroughly to which parts of the images people looks and base their evaluation.

The results are important in understanding the perception and identification of military visual objects in displays and can be used for example in optimization of decoys as well as, in connection with detection, display settings.