Utilization of Color Similarity Index for Evaluating …Yogi Tri Prasetyo School of Industrial...

Proceedings of the International Conference on Industrial Engineering and Operations Management

Dubai, UAE, March 10-12, 2020

© IEOM Society International

Utilization of Color Similarity Index for Evaluating Existing

Military Camouflage Designs

Yogi Tri Prasetyo

School of Industrial Engineering and Engineering Management

Mapúa University

658 Muralla St., Intramuros, Manila 1002, Philippines

[email protected]

Abstract

Military camouflage is an important part of defense technology. It is designed to confuse the enemy by visually

merging the outline of military design to the surrounding environment. The purpose of this study was to apply a color

similarity index for evaluating existing military camouflages designs. Camouflage Similarity Index (CSI) was utilized

as a color similarity index and the value varies between 0 to 1. The best value of 0 is achieved if the selected existing

military camouflage design perfectly blends with the surrounding environment. 10 existing military camouflage

designs from different regions were evaluated under 14 different locations in the swamp environment. The results

indicated that the CSI was an effective tool for identifying the effectiveness of existing military camouflage designs

across regions. Interestingly, even the CSI values were found different among 10 selected designs, Post-hoc Tukey

HSD test revealed that there was statistical difference between each design and it could be categorized into 3 different

groups. This study contributed to the advancement of color similarity index to the existing military camouflage and the results would be very useful for military research organizations, ministry of defense, and textile engineer.

Keywords

Color Similarity Index, Military Camouflage, Camouflage Similarity Index, Defense Technology, Color Algorithm.

1. Introduction

Military camouflage is an important part of the army combat uniform. It is an attempt to minimize the difference

between the army combat uniform and the surrounding background so that human eyes and military detection

instruments struggle to detect and distinguish the target (Xue et al., 2016). Military research organizations and color

researchers are continually engaged in the test and evaluation of prospective camouflage patterns, seeking to maximize

concealment while also considering the background heterogeneity of diverse operational contexts (Brunye et al., 2017;

Chang et al., 2012; Lin et al., 2014c; Xue et al., 2016b; Xue et al., 2018). By advancing military camouflage research,

soldiers could improve the survivability and mission effectiveness by preventing visual observation and other military

sensors from detecting both the soldiers and their equipment (Killian & Hepfinger, 1992; Chang et al., 2012;

Fincannon et al., 2013).

Military camouflage effectiveness is often assessed by image quality assessment algorithm. Previously, Zhang et al

(2013) proposed an algorithm based on Fourier transform and Gaussian low-pass filter (LPF) to mix the color based

on tricolor angular frequencies. In the model, the tricolor angular frequencies were introduced to the spatial frequency

response function of human color vision, and the effects of atmospheric attenuation and air screen brightness on color

mixture were also considered (Zhang et al., 2013). The field test indicated that the model can simulate the color-

mixing process in the aspects of the color-mixing order, and shape and position of the color-mixing spot. However,

the color-mixing spot was found not perfect. Xue et al (2016a) extracted primary colors from the background using a

k-means clustering algorithm to generate the color constraints. In addition, a spot template distribution algorithm was

proposed to generate camouflage patterns. Even this study achieved good results in terms of optical camouflage,

however, proposing this method to enhance an existing military camouflage would be difficult since it would change

totally the currently existing design.

1830

mailto:[email protected]




A computational approach using an image quality assessment algorithm is therefore helpful in overcoming the

limitations of assessing camouflage when using human observers (Le et al., 2019; Goudarzi et al., 2012; Volonakis et

al., 2018; Yang et al., 2019). Previously, we developed Camouflage Similarity Index (CSI) to access the effectiveness

of selected military camouflage (Lin et al., 2014b). This image quality assessment algorithm was found outperformed

commonly used image quality algorithms such as UIQI (Lin et al., 2014b), MSE, PSNR, and SSIM (Lin et al., 2014a;

Patil & Pawar, 2017; More & Borse, 2017).

The purpose of this study was to apply a color similarity index for evaluating existing military camouflages designs.

The results of this study could be used for the improvement of the selected existing military camouflages. This study

contributed to the advancement of color similarity index to the existing military camouflage and the results would be

very useful for military research organizations, ministry of defense, and textile engineer.

2. Methodology

2.1 Existing Military Camouflages Collection

10 different existing camouflages from 10 different countries were selected in the current study. The camouflages were

obtained using Google search engine by typing “(name of country) camouflage” or “(name of country) military

uniform”. The image later was cut to 20x50 pixels using Adobe Photoshop CS6 for CSI calculation (Figure 1). Matlab

2018 was used to calculate overall L*, a*, and b* values of each camouflage (Table 1).

Figure 1. Adobe Photoshop CS6 was used to cut the image to 20x50 pixel (Lin et al., 2019b; Lin et al., 2019c)

1831




Table 1. Overall L*, a*, and b* values of each camouflage

1

2

3

4

5

6

7

8

9

10

L* 75.3027

59.8683 58.0483 71.0352

55.7600 61.7693 70.3756 73.2909 55.0541 69.1774

a* -1.3365

-4.3183 1.6605 -1.0440

-4.3076 -10.3360 -2.0521 -18.6728 -2.4078 -0.1778

b* 9.4917 6.8219 17.2644 16.9246 -1.4285 10.3662 2.9419 13.9534 10.1500 13.3228

2.2 Background Collection

Considering different terrains and different climatic conditions, camouflaging is a challenging task for defense

applications (Karpagam et al., 2016; Brunye et al., 2018). At the moment, one camouflage is impossible to be applied

in all different terrains (Sparks, 2012). In the current study, one woodland background was selected. Canon EOS 5D

Mark II was used to capture the woodland background at 09:00 am (Figure 2).

Figure 2. Selected woodland background (Lin et al., 2019b; Lin et al., 2019c; Prasetyo, 2019)

14 different locations (backgrounds) from the woodland background were selected in the current study. 14 different

locations were selected to evaluate the camouflage effectiveness as a demonstration of evaluation in different

woodland environments since each location had different L*, a*, and b* values. Similar to camouflage, the image

location was cut to 20x50 pixels using Adobe Photoshop CS6 for CSI calculation. Matlab 2018 was also used to

calculate overall L*, a*, and b* values of each camouflage (Table 2).

Table 2. Overall L*, a*, and b* values of each background

1 2 3 4 5 6 7 8 9 10 11 12 13 14

L* 70.9752 76.1949

41.7265 47.4693 59.7803 59.3054 64.1585 52.4517 71.0048 67.8591 56.0665 51.3341 62.1479 78.3972

a* -4.7106 -4.1246

-4.0503 -0.8253 -7.8139 -7.8107 -8.5197 -7.8344 -5.9997 -7.9443 -6.3388 -6.3474 -4.1854 -7.2132

b* 12.4265 10.6737 2.2426 -1.1642 14.5979 14.5174 15.4299 13.2100 14.0740 15.7666 12.6836 9.3048 11.0375 16.9961

1832




2.3 Color Similarity Index

Figure 3. CSI calculation chart (Lin et al., 2014b; Lin et al., 2019b; Lin et al., 2019c; Prasetyo, 2019)

2.4 Statistical Analysis

Minitab 18 was used to calculate the significant difference between the CSI value. Post-hoc analysis among ten

selected military camouflage was performed. The results were considered statistically significant when p ≤ 0.05 (Lin

et al., 2018; Lin et al., 2019a; Martinez et al., 2019; Miraja et al., 2019; Prasetyo et al., 2014; Prasetyo et al., 2019;

Torres et al., 2019).

RGB information collection

in matlab

RGB information collection

in matlab

Conversion to XYZ space in

matlab

Conversion to XYZ space in

matlab

Conversion to CIELAB

space in matlab

Conversion to CIELAB

space in matlab

Calculate the CSI between

camouflage and selected

background in matlab

Repeat the measurement in

other backgrounds

Cut the image to 20 x 50

pixels in Adobe Photoshop

CS6

Cut the selected location

to 20 x 50 pixels in Adobe

Photoshop CS6

Find the camouflage using

Google search engine

Camouflage Background

Capture woodland

background using Canon

EOS 5D

1833




3. Results and Discussion

The purpose of this study was to investigate the camouflage effectiveness across selected countries using Camouflage

Similarity Index (CSI). CSI ranges from 0 to 1 and the best value 0 is achieved if camouflage perfectly blends with

the background.

The CSI results are presented in Table 3. Based on Table 3, design 1 had the lowest CSI on background 1. Design 4

had the lowest CSI on background 1 and 14, design 5 had the lowest CSI on background 3 and 4, design 6 had the

lowest CSI on background 5,6,7,8,10,11,12,13, and design 10 had the lowest on background 9. Regarding the highest

CSI on each background, design 1 had the highest CSI on background 1, design 5 had the highest CSI on background

1,2,5,6,7,8,9,10,11,12,13, and design 8 had the highest CSI on background 3,4,12,13.

Table 3. Camouflage Similarity Index (CSI) results Background Camouflage Similarity Index (CSI)

1

2

3

4

5

6

7

8

9

10

Background 1 0.6883 0.8428 0.8275 0.6768** 0.9639* 0.7044 0.8623 0.8687 0.7975 0.7226

Background 2 0.6725** 0.8932 0.8855 0.7906 0.9605* 0.8384 0.7761 0.9009 0.8668 0.8026

Background 3 0.9842 0.9052 0.9464 0.9762 0.7823** 0.9406 0.9434 0.9931* 0.9155 0.9699

Background 4 0.9823 0.8615 0.9581 0.9826 0.6523** 0.9638 0.9237 0.9962* 0.9483 0.9732

Background 5 0.8857 0.8663 0.9085 0.8288 0.9643* 0.7638** 0.9157 0.8908 0.8664 0.8713

Background 6 0.8928 0.8637 0.9095 0.8331 0.9631* 0.7605** 0.9099 0.8911 0.8670 0.8781

Background 7 0.8784 0.8765 0.8922 0.7930 0.9517* 0.6964** 0.9101 0.8513 0.8118 0.8693

Background 8 0.9468* 0.8954 0.9210 0.9244 0.9393 0.7977** 0.9446 0.9449 0.8647 0.9039

Background 9 0.7892 0.9049 0.8943 0.8100 0.9554* 0.8238 0.8924 0.8945 0.8449 0.7715**

Background 10 0.8300 0.8341 0.8855 0.7425 0.9614* 0.6854** 0.9058 0.8519 0.8090 0.7989

Background 11 0.9120 0.9162 0.9305 0.8804 0.9472* 0.8448** 0.9457 0.9129 0.8932 0.8854

Background 12 0.9368 0.8554 0.9008 0.9107 0.8322 0.7330** 0.9025 0.9586* 0.8531 0.9135

Background 13 0.8031 0.8342 0.8557 0.7817 0.9025 0.7686** 0.8856 0.9068* 0.8514 0.8287

Background 14 0.7603 0.9037 0.8834 0.7250** 0.9900* 0.8859 0.9123 0.7904 0.8876 0.8320

Average 0.8545 0.8752 0.8999 0.8326 0.9119 0.8005 0.9022 0.9037 0.8627 0.8586

Max 0.9842 0.9162 0.9581 0.9826 0.9900 0.9638 0.9457 0.9962 0.9483 0.9732

Min 0.6725 0.8341 0.8275 0.6768 0.6523 0.6854 0.7761 0.7904 0.7975 0.7226

Regarding the overall performance, design “6” was found had the lowest average CSI (0.8005) and design “5” had

the highest average CSI (0.9119) (Figure 4). However, even design “6” had the lowest CSI, this design was found not

suitable to be applied in dark woodland environment such as backgrounds 3 and 4. In these backgrounds, design “5”

was found had the lowest CSI with the value of 0.7823 and 0.6523 respectively.

One-way ANOVA was applied to test the significance of the design to CSI. Based on Table 4, it was that there was a

significant effect of design to the CSI. Tukey HSD test was applied to test the significance of multiple comparisons.

This test can simultaneously run the set of all pairwise comparisons and identifies any difference between two means

that is greater than the expected standard error (Table 5). Interestingly, there was 3 different groups which consist of

group A (design 6), group B (4,1,10,9,2), and group C (3,7,8,5).

1834




Figure 4. Interval Plot of CSI vs Design.

Table 4. ANOVA for Design vs CSI.

Source DF Adj SS Adj MS F-Value P-Value

Design 9 0.1610 0.017891 3.66 0.000

Error 130 0.6357 0.004890

Total 139 0.7967

Table 5. Means, SDs, and Tukey HSD test result of “S” country

Design Mean StDev Group

6 0.8005 0.0874 A

4 0.8326 0.0921 AB

1 0.8545 0.1007 AB

10 0.8586 0.0714 AB

9 0.8627 0.0411 AB

2 0.8752 0.0280 AB

3 0.8999 0.0341 B

7 0.9022 0.0431 B

8 0.9037 0.0563 B

5 0.9119 0.0941 B

Despite the substantial and clear study results, the authors would like to acknowledge the limitations of the current

study. First, the lack of proper military camouflage collection. Instead of using the proper design, we obtained the

camouflages by using the Google search engine. This would lead to unbalance color distribution. Moreover, the

sunlight or when the picture was taken would definitely affect the results. Second, the CSI values were limited to the

selected woodland background which was taken at 09:00 am. Different sunlight, terrains, and thermal conditions

would influence the environment which subsequently influences the CSI values. Future research to assess the selected

camouflage in dessert or other terrains would be a promising research topic.

4. Conclusions

Military camouflage is an important part of defense technology. It is designed to confuse the enemy by visually

merging the outline of military design to the surrounding environment. The purpose of this study was to apply a color

similarity index for evaluating existing military camouflages designs. Camouflage Similarity Index (CSI) was utilized

1835




as a color similarity index and the value varies between 0 to 1. The best value of 0 is achieved if the selected existing

military camouflage design perfectly blends with the surrounding environment. 10 existing military camouflage

designs from different regions were evaluated under 14 different locations in the swamp environment. The results

indicated that the CSI was an effective tool for identifying the effectiveness of existing military camouflage designs

across regions. Interestingly, even the CSI values were found different among 10 selected designs, Post-hoc Tukey

HSD test revealed that there was statistical difference between each design and it could be categorized into 3 different groups. This study contributed to the advancement of color similarity index to the existing military camouflage and

the results would be very useful for military research organizations, ministry of defense, and textile engineer.

Acknowledgments

The authors would like to thank Nio Dolly Siswanto for his invaluable supports in this study.

References

Bacon, F. W., Iannarilli, F. J., Conant, J. A., Deas, T., & Dinning, M. (2009). Quantitative camouflage paint selection

for the CH-47F helicopter. Color Research & Application, vol.34(6), pp.406–416.

Brunye, T.T., Eddy, M.D., Cain, M.S., Hepfinger, L.B., Rock, K. (2017). Masked priming for the comparative

evaluation of camouflage conspicuity. Applied Ergonomics, vol.62, pp.259-267.

Brunye, T.T., Martis, S.B., Horner, C., Kirejczyk, J.A., Rock, K. (2018). Visual salience and biological motion interact

to determine camouflaged target detectability. Applied Ergonomics, vol.73, pp.1-6.

Chang, C-C., Lee, Y-H., Lin, C-J. (2012). Visual assessment of camouflaged targets with different background

similarities. Perceptual and Motor Skills, vol.114(2), pp.527-541.

Fairchild, M. (2015). Color models and systems. In A.Elliot, M.Fiarchild, A.Franklin (Eds.), Handbook of Color

Psychology (Cambridge Handbooks in Psychology, pp. 9-26). Cambridge: Cambridge University Press.

Fincannon, T., Keebler, J.R., Jenetsch, F., Curtis, M. (2013). The influence of camouflage, obstruction, familiarity

and spatial ability on target identification from an unmanned ground vehicle. Ergonomics, vol.56 (5), pp.739 –

751.

Goudarzi, U., Mokhtari, J., & Nouri, M. (2012). Camouflage of cotton fabrics in visible and NIR region using three

selected vat dyes. Color Research & Application, vol.39(2), pp.200–207.

Karpagam, K.R., Saranya, K.S., Gopinathan, J., Bhattacharyya, A. (2016). Development of smart clothing for military

applications using thermochromic colorants. The Journal of The Textile Institute, vol.108(7), pp.1122-1127.

Le, T.-N., Nguyen, T. V., Nie, Z., Tran, M.-T., & Sugimoto, A. (2019). Anabranch network for camouflaged object

segmentation. Computer Vision and Image Understanding, vol.184, pp.45–56.

Lin, C.J., Chang, C-C., Lee, Y-H. (2014a). Developing a similarity index for static camouflaged target detection. The

Imaging Science Journal, vol.62(6), pp.337-341.

Lin, C.J., Chang, C-C, Liu, B-S (2014b). Developing and evaluating a target-background similarity metric for

camouflage detection. PLoS ONE 9(2): e87310.

Lin, C.J., Chang, C-C., Lee, Y-H. (2014c). Evaluating camouflage design using eye movement data. Applied

Ergonomics, vol.45, pp.714-723.

Lin, C.J., Prasetyo, Y.T., Widyaningrum, R. (2018). Eye movement parameters for performance evaluation in

projection-based stereoscopic display. Journal of Eye Movement Research, vol.11(6):3.

Lin, C. J., Prasetyo, Y. T., & Widyaningrum, R. (2019a). Eye Movement Measures for Predicting Eye Gaze Accuracy

and Symptoms in 2D and 3D Displays. Displays, vol.60, pp.1-8.

Lin, C. J., & Prasetyo, Y. T. (2019b). A metaheuristic‐based approach to optimizing color design for military

camouflage using particle swarm optimization. Color Research & Application, vol.44(5), pp.740-748.

Lin, C.J., Prasetyo, Y.T., Siswanto, N.D., Jiang, B.C. (2019c). Optimization of color design for military camouflage

in CIELAB color space. Color Research & Application, vol.44(3), pp.367-380.

Martinez, J. E. F., Prasetyo, Y. T., Robielos, R. A. C., Panopio, M. M., Urlanda, A. A. C., & Topacio-Manalaysay, K.

A. C. (2019). The Usability of Metropolitan Manila Development Authority (MMDA) Mobile Traffic Navigator

as Perceived by Users in Quezon City and Mandaluyong City, Philippines. Proceedings of the 2019 5th

International Conference on Industrial and Business Engineering - ICIBE 2019, pp.207-211.

1836




Miraja, B. A., Persada, S. F., Prasetyo, Y. T., Belgiawan, P. F., & Redi, A. P. (2019). Applying Protection Motivation

Theory To Understand Generation Z Students Intention To Comply With Educational Software Anti Piracy

Law. International Journal of Emerging Technologies in Learning (IJET), vol.14(18), pp.39.

More, K., Borse, S.B. (2017). Camouflage texture assessment method based on WSSIM and nature. International

Journal of Engineering and Techniques, vol.3(3), May-June 2017.

Patil, K.V., Pawar, K.N. (2017). Method for improving camouflage image quality using texture analysis. International

Journal of Computer Applciations, vol.180 (8), pp.6-8.

Prasetyo, Y. T. (2019). Evaluating Existing China Military Camouflage Designs using Camouflage Similarity Index

(CSI). Proceedings of the 2019 5th International Conference on Industrial and Business Engineering - ICIBE

2019, pp.321-325.

Prasetyo, Y. T., Widyaningrum, R., & Lin, C. J. (2019). Eye Gaze Accuracy in the Projection-based Stereoscopic

Display as a Function of Number of Fixation, Eye Movement Time, and Parallax. 2019 IEEE International

Conference on Industrial Engineering and Engineering Management (IEEM).

Prasetyo, Y.T., Suzianti, A., Dewi, A.P. (2014). Consumer preference analysis on flute attributes in Indonesia using

conjoint analysis. International Conference on Advanced Design Research and Education (ICADRE), pp.111-117.

Sparks, E. (2012). Advance in Military Textile and Personal Equipments 1st edition. Cambridge, UK: Woodhead

Publishing.

Torres, M. E. S., Prasetyo, Y. T., Robielos, R. A. C., Domingo, C. V. Y., & Morada, M. C. (2019). The Effect of

Nutrition Labelling on Purchasing Decisions. Proceedings of the 2019 5th International Conference on Industrial

and Business Engineering - ICIBE 2019, pp.82-86.

Volonakis, T.N., Matthews, O.E., Liggins, E., Baddeley, R.J., Scott-Samuel, N.E., Cuthill, I.C. (2018). Camouflage

assessment: Machine and human. Computers in Industry, vol.99, pp.173-182.

Xue, F., Xu, S., Luo, Y-T., Jia, W. (2016a). Design of digital camouflage by recursive overlapping of pattern templates.

Neurocomputing, vol.172, pp.262-270.

Xue, F., Yong, C., Xu, S., Dong, H., Luo, Y., Jia, W. (2016b). Camouflage performance analysis and evaluation

framework based on features fusion. Multimedia Tools and Applications, vol.75, pp.4065-4082.

Yang, X., Xu, W.-D., Jia, Q., Li, L., Zhu, W.-N., Tian, J.-Y., & Xu, H. (2019). Research on extraction and reproduction

of deformation camouflage spot based on generative adversarial network model. Defence Technology, article in

press.

Zhang, Y., Xue, S-q, Jiang, X-j., Mu, J-y, Yi, Y. (2013). The spatial color mixing model of digital camouflage pattern.

Defense Technology, vol.9, pp.157-161.

Biography

Yogi Tri Prasetyo is currently an associate professor in the School of Industrial Engineering and Engineering

Management, Mapúa University, Philippines. He received a Bachelor of Engineering in Industrial Engineering from

Universitas Indonesia (2013). He also studied for one year (2011-2012) at Waseda University, Japan, during his junior

year as an undergraduate exchange student. He received an MBA (2015) and a Ph.D. (2019) from Department of

Industrial Management National Taiwan University of Science and Technology (NTUST), with a concentration in human factors and ergonomics. He was awarded as NTUST Outstanding Youth with a perfect GPA 4.00. He has a

wide range of research interests including human-computer interaction particularly related to eye movement, color

optimization of military camouflage, strategic product design, usability analysis, and now he is currently doing

accident analysis and prevention. He published several SCI journals in Displays, Color Research and Application,

Journal of Eye Movement Research, several non-SCI journals, and several conference proceedings. In addition,

Dr.Yogi has contributed to several international conferences as co-chair, chair session, and even committee members.

Apart from academics, Dr.Yogi likes playing flute, judo, swimming, and hiking. He has two black belts (1st dan black

belt judo and 1st dan black belt taekwondo), an international certified lifeguard, and a certified advanced diver. He

speaks Indonesian, English, Chinese, Japanese, and currently, he’s working hard for his Filipino.

1837

Utilization of Color Similarity Index for Evaluating …Yogi Tri Prasetyo School of Industrial...

Documents

Transcript of Utilization of Color Similarity Index for Evaluating …Yogi Tri Prasetyo School of Industrial...