Journal of Physics: Conference Series

PAPER • OPEN ACCESS You may also like

- Futurity scenario of water resources
Using Optimum Index Factor and Determinant management in Bahr An-Najaf basin in
Iraq
Covariance Methods and Compare to PCA on Hassan Ali Omran, Mohammed Sh
Mahmood and Ali Abbas Kadhem

Satellite Images to Determine The Earth’s - Challenges and opportunities of the urban
structures of holy cities: the case of old
Landmarks ( Part of Tar- An Najaf and Its Najaf, Iraq
Sabeeh Lafta Farhan, Ula Abd Ali Khalel
Neighbours) Merie and Dhirgham Alobaydi

- The role of street and park trees to CO2

removal and improving air quality and
To cite this article: Sa’ad R Yousif and Watheq F Shneen 2020 J. Phys.: Conf. Ser. 1660 012077 climate in urban areas in Najaf (Iraq)Type
the title of your paper here
Muthik A. Guda, Jawad K. Obaid, Faris j.
Alduhaidahawi et al.

View the article online for updates and enhancements.

This content was downloaded from IP address 37.239.86.46 on 17/07/2023 at 23:46

1st International Conference on Pure Science (ISCPS-2020) IOP Publishing
Journal of Physics: Conference Series 1660 (2020) 012077 doi:10.1088/1742-6596/1660/1/012077

Using Optimum Index Factor and Determinant Covariance

Methods and Compare to PCA on Satellite Images to Determine The
Earth’s Landmarks ( Part of Tar- An Najaf and Its Neighbours)

Dr. Sa'ad R Yousif 1 2,Watheq F Shneen2.

1
Univ. of Kufa/ Faculty of Science/ Dept. of Geology.
2
Univ. of Kufa/ Remote Sensing Center.

Abstract. The research focuses on mixing of Landsat8 ETM+ images for sector of Tar An- Najaf
Phenomenon for the sake of gaining the optimum description showing the ground guidelines of
the studied area. The optimum Index Factor (OIF) method has ensured that RGB567 represents
the preferred mixed bands, where vegetated areas, water and rocks and deposits more obviously
distributed. The determinant covariance matrix method stated that RGB457 considered the
optimum for spectral resolution. For comparison with Principal Component Analysis, the latter
involves PC1,2,3 and shows clear variations that serve the interpreter as a detailed study.
Through the study, the best method is the second one.

The Aim of the Study

The study aims to obtain the excellent representation of earth phenomena by composing RGB bands of
the colour composite satellite image of Landsat 8 ETM+ using numerical techniques, as well as knowing
which of the two methods are the best.

1. Introduction
Images taken by remote sensing satellites in their unprocessed status are often ineffective in detection
and accurate perception. Digital processing methods are used to enhance the spectral interpretation;
therefore, the interpreter can deduce these concepts so accurately. The intent of image processing
approaches is to enhance and focus the spectral utility of terrestrial submissions [1]. The images are
stored in a three-dimensional matrix. The three dimensions of the matrix represent the intensities of the
red, green, and blue bands. That depiction of image as an arrangement enables to work with the image
softly (i.e. calculations and treatments can be performed [2].
The most typical image processing executed on these modes is the composition of the RGB image
of the most interpretable bands of the same parameter, for example the method of calculating the
optimum index factor (OIF) [3]. Through these two factors, the lowest correlation bands are obtained
[4]. OIF is a statistical value that can be used to determine the optimal (best) combination of three
bands in the satellite image in which the ability to produce a clear perception can be done [5], whilst
the determinant matrix covariance is used to calculate the determinant of a matrix consisting of three
dimensional representations of bands. The sensors are sensitive to electromagnetic spectrum resulting
images different in their resolutions due to their sensors. So, Landsat8 ETM+ is used for its latest
launching and effectiveness.

Content from this work may be used under the terms of the Creative Commons Attribution 3.0 licence. Any further distribution
of this work must maintain attribution to the author(s) and the title of the work, journal citation and DOI.
Published under licence by IOP Publishing Ltd 1
1st International Conference on Pure Science (ISCPS-2020) IOP Publishing
Journal of Physics: Conference Series 1660 (2020) 012077 doi:10.1088/1742-6596/1660/1/012077

2. Location of the Studied Area

The research focused on sector of Tar An-Najaf and the surrounding area located in An-Najaf Al-Ashraf
Province. Data were taken from Landsat satellite 8 ETM + (LC81680382014130-01014-05-10) from
p168-r38, where the area lies within (43 ° 56' 28.88'' E, 32 ° 7' 48.86'' N & 32 ° 2' 8.42''N 44 ° 10'
54.06'' E), Figure (1). Tar An-Najaf is a cliff that differs in declination due to various geological factors,
and As-Sadeer (Al-Ghazi) River, which is one of the small branches of the Euphrates River runs along
the side of the escarpment.

Figure 1. The location of studied area within Iraq Source work researcher.

3. Methodology
The programs used in the research are:
1- ERDAS IMAGINE 2013 for the sake of image processing (by means of which the satellite
bands are exported, and the intrinsic value relationship is applied in the way the basic
components are converted and read in the resolution of 15m.
2- ENVI 4.7 also for the sake of handling image processing
Enhancement methods aim to make visual information clearer for visual interpretation purposes.
Raw visual information is often ill-prepared in a way that is appropriate to the sensitivity of the human
eye. In addition, enhancement methods sometimes aim to build a physical link between brightness
variance amongst multiple bands of regular features [5]. So, the image processing can emerge the
required information efficiently to the interpreters by the following techniques used in the research.

1- Arrangement of the RGB image

A- Finding the optimum index factor (OIF) from the equation [2] and [6]:

2
1st International Conference on Pure Science (ISCPS-2020) IOP Publishing
Journal of Physics: Conference Series 1660 (2020) 012077 doi:10.1088/1742-6596/1660/1/012077

∑ 
(1) OIF =  ∑ 

The correlation coefficient that was extracted using ERDAS IMAGINE 8.4 is derived using the
Covariance ratio.
B- Obtaining the covariance matrix determinant by the following formula of The Sheffield index
(SI), [7]
(2) SI = DET(B )
Where Bxyz is defined as bxy, bxz ............, bzy the two bands discrepancy values of the three bands
(3) Lsi = LOG (SI) (by experiment)

2- Special processing are the mode of conversion of principal component analysis, where the
extraction of the intrinsic value (Eigenvalue), which the one of them has three self-vectors
(eigenvectors). The first intrinsic values are called the first PC1 and the other values are PC2, PC3
...... and can be simplified as in equation:
(4) PC 123 = Log (pc 123)

4. Application
The emergence of vegetative areas near the swamps and ponds is obscured through the viewing of true
colour image of the area, Figure (2)

Figure 2. Landsat 8 True colour image of the study area on 2014-05-10 using the ENVI program.
1- Composition of composite image
A- Optimum Index Factor OIF
It can be found for every RGB bands from equation (1).
Table 1. OIF values for different RGB bands of the studied area.

bands OIF Bands OIF

56 7 4751 146 3205

46 7 4477 345 3016
36 7 4202 237 2926
457 4193 236 2897
456 4119 137 2859

3
1st International Conference on Pure Science (ISCPS-2020) IOP Publishing
Journal of Physics: Conference Series 1660 (2020) 012077 doi:10.1088/1742-6596/1660/1/012077

26 7 4069 136 2826

16 7 4036 245 2821
357 3880 145 2744
356 3808 127 2638
257 3713 126 2605
157 3658 235 2437
256 3631 135 2353
156 3570 125 2122
347 3474 234 2048
346 3442 134 1967
247 3309 124 1751
246 3268 123 1360
147 3251

Figure 3. colour composite image RGB567 of Landsat 8 ETM + dated 2014-05-10 .

B- Calculating Covariance matrix determinant separately for any RGB bands
Table 2. The values of Covariance matrix determinant of the studied area in a descending order .
LOG(det) BANDS LOG(det) BANDS

19.801 457 18.013 146

19.572 357 17.988 145
19.389 456 17.906 345
19.283 257 17.898 346
19.268 467 17.767 137
19.193 356 17.714 237
19.173 567 17.643 136
19.099 157 17.580 236
19.083 367 17.447 135
18.921 256 17.367 235

4
1st International Conference on Pure Science (ISCPS-2020) IOP Publishing
Journal of Physics: Conference Series 1660 (2020) 012077 doi:10.1088/1742-6596/1660/1/012077

18.784 267 16.524 127

18.748 156 16.399 126
18.608 167 16.217 125
18.267 247 16.207 134
18.215 157 16.134 234
18.199 347 15.186 124
18.051 246 14.275 123
18.026 245
From Table (2) any RGB band value can be seen distinctly. Of bands 4, 5 and 7, the relation can be
illustrated as follows, Table (3)
Table 3. The interrelation of Covariance matrix standards for bands 4, 5 and 7.
Covariance Band 4 Band 5 Band 7
Band 4 8793095.897 10400939.22 12091176.25
Band 5 10400939.22 14020351.22 15168464.68
Band 7 12091176.25 15168464.68 21253267.65

Then the own determinant can be extracted using Equation (3) as follows:
LOG (DETM∗)= 19.801
The highest values were for RGB 457 in Figs. 4 and 357 in Fig. 5 without the thermal band, because
the amount of variance of bands 3 and 4 with the remaining bands was the highest between the bands

Figure 4. Determination of covariance matrix determinant of RGB 457 of Landsat 8 dated 2014-05-10
for the study area.

5
1st International Conference on Pure Science (ISCPS-2020) IOP Publishing
Journal of Physics: Conference Series 1660 (2020) 012077 doi:10.1088/1742-6596/1660/1/012077

Figure 5. Determination of covariance matrix determinant of RGB 357 of Landsat 8 dated 2014-05-
10 for the study area.

Figure 6. Determination of covariance matrix determinant of RGB 456 of Landsat 8 dated

2014-05-10 for the study area.
Figures 4,5, & 6 are similar in values and logarithms, so the images are very close to the sight.
2- Finding the square root of the multiplication of the intrinsic values corresponding to separate
band.

6
1st International Conference on Pure Science (ISCPS-2020) IOP Publishing
Journal of Physics: Conference Series 1660 (2020) 012077 doi:10.1088/1742-6596/1660/1/012077

Table 4. Eigenvalues of PCA arranged in a descending order for the higher ten values.

Mix Log(Eigen)

123 20.450

124 19.470

125 19.378

126 19.129

127 18.500

128 17.248

129 16.553

134 19.035

135 18.944

Figure 7. Mixing of the principal components 123 which is the best mix using Erdas program.
Table (4) shows the composition of the principal components 123 as in Figure (7) where the green
color refers to the extensions of Bahr An-Najaf Depression and the surrounding drainage basin of
artesian wells as well as As-Sadeer River and the ponds it consists of. The blue color refers to the wet
soil that used as cultivated and agricultural areas, while the light blue color indicates the saturation with
water. The variance in the carroty and ruddy colors indicates the soil pattern. That is because the study

7
1st International Conference on Pure Science (ISCPS-2020) IOP Publishing
Journal of Physics: Conference Series 1660 (2020) 012077 doi:10.1088/1742-6596/1660/1/012077

area represents a prominent escarpment, and the surroundings are lowlands. The prominent dark pink
colour that outline Tar An-Najaf represents part of Wadi As-Salam Cemetery.
The standards of the covariance array stay unchanged during the conversion. The sum of the
variations of the image before and after the conversion leading to the constant data of the image. In the
covariance matrix of the pre-conversion, the logarithm of the total sum of the diagonal elements of the
matrix (Table 5) equals 7.90, and logarithm
Table 5. Covariance standards of the image denoting to the studied area before conversion.
Covariance Band 1 Band 2 Band 3 Band Band 5 Band 6 Band 7 Band 8 Band 9
Band 1 890873.3633 1158115.8 1723039.877 2694156 3145136.973 4065691.527 3364019.69 2054174.5 3170.984
Band 2 1158115.849 1513305.8 2269078.024 3559776.8 4153404.213 5429347.49 4543546.02 2708953.2 4356.4428
Band 3 1723039.877 2269078 3469049.23 5475452.7 6373286.128 8467043.154 7220573.53 4148149.58 7171.2303
Band 4 2694155.952 3559776.8 5475452.685 8793095.9 10400939.22 13997065.51 12091176.3 6604537.33 13165.442
Band 5 3145136.973 4153404.2 6373286.128 10400939 14020351.22 17821759.78 15168464.7 7752403.15 18329.839
Band 6 4065691.527 5429347.5 8467043.154 13997066 17821759.78 24833879.79 22362385.7 10358365.3 26614.171
Band 7 3364019.689 4543546 7220573.527 12091176 15168464.68 22362385.73 21253267.7 8888046.82 25125.105
Band 8 2054174.496 2708953.2 4148149.58 6604537.3 7752403.148 10358365.26 8888046.82 5140094.59 9263.8066
Band 9 3170.983976 4356.4428 7171.230299 13165.442 18329.83891 26614.17094 25125.1048 9263.80664 187.77581
of the sum of the intrinsic values of all bands in the covariance matrix after conversion (Table 6) is
7.90 indicating that the standards of the covariance stay the same.
Table 6. Covariance standards of the image denoting to the studied area after conversion.
Eigenvalue pc 1 pc 2 pc 3 pc 4 pc 5 pc 6 pc 7 pc 8 pc 9
pc 1 75258329.08 0.001611 -0.002066 0.000128 0.000941 0.000047 0.000682 -0.002064 -0.001514
pc 2 0.001611 3190654.316 0.000551 0.000461 0.000754 0.000232 0.000605 -0.000543 -0.000468
pc 3 -0.002066 0.000551 1172522.722 -0.00006 0 0.000069 0.000017 0.000755 0.000338
pc 4 0.000128 0.000461 -0.00006 122937.0244 -0.543819 -0.000031 0.000101 -0.000163 -0.000199
pc 5 0.000941 0.000754 0 -0.543819 99527.45855 -0.000034 0.00023 -0.000254 -0.000391
pc 6 0.000047 0.000232 0.000069 -0.000031 -0.000034 56088.37098 -0.000061 0.000136 0.000015
pc 7 0.000682 0.000605 0.000017 0.000101 0.00023 -0.000061 13160.4401 0.000069 -0.000002
pc 8 -0.002064 -0.000543 0.000755 -0.000163 -0.000254 0.000136 0.000069 737.027016 -0.000005
pc 9 -0.001514 -0.000468 0.000338 -0.000199 -0.000391 0.000015 -0.000002 -0.000005 148.836947
The distribution of the band before conversion is unevenly, while the information are largely
concentrated in the first component after conversion, and the rest is distributed on the second and third
and so on, Table (7).
Table 7. Percentage of covariance before and after conversion
Post conversion component Pre-conversion Band
94.17402% PC1 10.54% B1
3.99260% PC2 10.94% B2
1.46723% PC3 11.58% B3
0.15384% PC4 12.30% B4
0.12454% PC5 12.66% B5
0.07019% PC6 13.10% B6
0.01647% PC7 12.98% B7
0.00092% PC8 11.88% B8
0.00019% PC9 4.03% B9
100% 100% sum

8
1st International Conference on Pure Science (ISCPS-2020) IOP Publishing
Journal of Physics: Conference Series 1660 (2020) 012077 doi:10.1088/1742-6596/1660/1/012077

5. Discussion and Conclusion

The research clarifies that the OIF and Variance Matrix Determinant techniques depend on a numerical
process without adjustments in the unique data of the image. The second method is more suitable than
the first one. While the method of conversion of the principal components are computational
calculations, and introduce conversions to modify the original values without any effect on the variance
standards before the conversion
The latter shows pseudocolors to clarify extra features in the processed image, and enables
interpreters to get better the output and decision making, and thus it considers the optimum method than
the the first two ones.

References
[1] Al-Anqari, K.M., 1986, "Remote Sensing and its Applications in Spatial Studies", Dar Al-Marikh
Publishing, Riyadh, 179p.
[2] Al-Juaidi, F., Millington, C.A., & Mclaren, J.S., 2003, "Merged remotely sensed data for
geomorphological investigations in deserts: examples from central Saudi Arabia", The
Geographical Journal, pp (117–130).
[3] Chavez, P.S., Berlin, G.L. & Sowers, L. B., 1982, " Statistical method for selecting Landsat MSS
ratios", Journal of applied photographic engineering, 8(1), pp (23-30).
[4] Richards, A.J. and Jia, X., 2006, "Remote Sensing Digital Image Analysis an Introduction",4th
Edition, Canberra, Australia, 453p.
[5] Haque, M., 2011," Selecting optimum band combinations for the visualisation of eight-band
Worldview 2 data", PhD Researcher, Monash University, Australia, Pp(3-4),
Email:[email protected].
[6] Patel, N. and Kaushal, B., 2011," Classification of features selected through Optimum Index
Factor (OIF) for improving classification accuracy", Journal of Forestry Research 22(1): pp
(99−105).
[7] Sheffield, C., 1985, "Selecting band combinations from multispectral data", Photogrammetric
engineering and remote sensing, pp (51, 681-687).