Professional Documents
Culture Documents
Research Paper
a r t i c l e i n f o a b s t r a c t
Article history: Image classification of satellite imagery interprets the thematic map to represent the spatial distribution
Received 13 November 2017 of earth features. There are so many applications of Remote sensing image classification such as Resource
Revised 31 March 2018 utilization and environmental impact analysis etc. The overall process result depends on two main
Accepted 26 April 2018
aspects (1) Every object have distinctive signature and feature of interest (2) The process can distinguish
Available online 18 December 2018
these features separately. Image classification is broadly classified in two ways (1) Hard classification and
(2) Soft classification. In hard classification, pixels are classified in to a single class only and in soft clas-
Keywords:
sification, pixels can belong to one or more classes according to their membership values. In hard classi-
Hard classification
Soft classification Fuzzy C-Means
fication, data may lost because of the restriction being in a single class only. But in soft classification, this
Accuracy problem is resolved. But after resolving the problem, there is a need of accuracy assessment. There are
Entropy many commercial software available in market but they are not providing accuracy assessment for soft
classified images. So, in this study, a tool is designed to overcome such a problem.
Ó 2018 National Authority for Remote Sensing and Space Sciences. Production and hosting by Elsevier
B. V. This is an open access article under the CC BY-NC-ND license (http://creativecommons.org/licenses/by-
nc-nd/4.0/).
1. Introduction a cluster, whose identity is not known. Then this cluster is con-
verted into information classes by the analyst. Thus the definition
Remote sensing is a science by which we gather the information of the cluster is based on some spectral property of the object
of the surface of the earth. But those image or data captured by hence it is known as spectral class (See Tables 1, 2, 3, 4, 5, 6, 7, 8).
satellite belong to the multi-classes. In terms of hard classification, The remote sensing image includes a combination of varied and
it supports a data which belong to a single class. But there is some pure pixels. In digital image taxonomy, image or a pixel is habitu-
amount of data loss in hard classification, (Bora and Gupta, 2014) ally measured like a unit belongs to only land cover category. How-
and this problem is overcome by using soft classification. The pro- ever, due to constricted image resolution, pixel often represents
cess of multispectral satellite classification (Wojtaszek et al., 2015) ground area, which consists of more than single isolated land cover
may be performed by any of the two approaches namely; super- classes. For this cause, it has been considered that fuzziness should
vised and unsupervised techniques. In supervised classification be accommodated in the categorization technique so that pixels
approach (Tamouk et al., 2013; Stein, 2002) (Kalra et al., 2013), may have multiple or partial class membership (Foody, 1996). In
the analyst classifies the image on the basis of known information this case, a determine of the strong point of membership or rela-
referred to as training data. This stage is also called signature anal- tionship for every class is output by the classifier, resultant in a soft
ysis. In supervised classification, training data plays a vital role. It classification technique (Yannis and Stefanos, 1999). A unsurpris-
provides a blue print on which a classifier assigns a pixel to its ing ‘hard’ classification technique, which allocates each pixel to a
information class. In unsupervised classification, it is based on explicit class, is often unsuitable for application where miscella-
some natural spectral properties or some parameter, i.e. (Dwivedi neous pixels are profuse in the image (Jabari and Zhang, 2013).
et al., 2012) (Dwivedi et al., 2012a,b,c), the image is converted into Some of the commercially available digital image processing
software, such as, Environment for Visualizing Images (ENVI),
Earth Resource Data Analysis System (ERDAS), Earth Resource
Peer review under responsibility of National Authority for Remote Sensing and
Space Sciences. Mapping software (ER Mapper) and IDRISI do not provide any cor-
⇑ Corresponding author. responding accuracy measures for soft classified output of their
E-mail addresses: sharmaranjana04@gmail.com (R. Sharma), principal.computers@ evaluation. So, in this study, a tool is designed to overcome such
tmu.ac.in (R.K. Dwivedi).
https://doi.org/10.1016/j.ejrs.2018.04.007
1110-9823/Ó 2018 National Authority for Remote Sensing and Space Sciences. Production and hosting by Elsevier B. V.
This is an open access article under the CC BY-NC-ND license (http://creativecommons.org/licenses/by-nc-nd/4.0/).
50 R. Sharma et al. / Egypt. J. Remote Sensing Space Sci. 23 (2020) 49–55
Table 1
AWIFS entropy of various land cover classes fromFCM classification output.
Value of weighting Agriculture bright forest dense forest Agriculture dry Agriculture moist Water Body
exponent ‘m’ land without crop land without crop
1.1 0 0 0 0 0 0
1.3 0 0 0.495736 0 0 0
1.5 0.005647 0.720945 0.534943 0.849842 0.005647 0
1.7 0.005647 0.940597 0.55884 0.917775 0.005647 0
1.9 0.005647 0.946659 0.595714 0.93509 0.005647 0
2.1 0.005647 0.964547 0.602791 0.93509 0.005647 0.005647
2.3 1.38439 1.019735 0.702469 0.962081 0.577753 0.005647
2.5 1.387307 1.023428 0.900355 0.968725 0.972232 0.005647
2.7 1.422581 1.024072 0.995434 0.989274 0.974703 0.005647
2.9 1.465228 1.16432 0.996447 0.995646 0.995284 0.005647
3.1 1.493104 1.184123 1.066264 0.998234 1.053199 0.005647
3.3 1.493358 1.205585 1.095643 1.008455 1.073952 0.569343
3.5 1.512936 1.367753 1.103139 1.058969 1.111597 0.766624
3.7 1.519854 1.376267 1.119413 1.071822 1.11494 0.892742
3.9 1.612968 1.438993 1.130291 1.13704 1.117274 1.013136
4.1 1.663884 1.445044 1.130666 1.170324 1.150236 1.273812
4.3 1.691301 1.462781 1.133814 1.195284 1.151883 1.348145
4.5 1.698294 1.536436 1.143974 1.22129 1.204915 1.395559
4.7 1.746609 1.652529 1.199743 1.372729 1.246343 1.404513
4.9 1.874828 1.742236 1.469495 1.422326 1.248483 1.474453
Table 2
AWIFS entropy of various land cover classes from PCM classification output.
Value of weighting Agriculture bright forest dense forest Agriculture dry land Agriculture moist Water Body
exponent ‘m’ without crop crop land without crop
1.1 0.005647 0.36168 0.016873 0.279126 0.108342 0
1.3 0.005647 0.736859 0.512143 1.767634 0.148719 0
1.5 1.008899 1.367815 1.447699 1.889342 0.351796 0
1.7 1.332126 1.913278 1.612615 1.897806 1.045161 0.351796
1.9 1.793768 2.129041 2.208572 1.920896 1.703991 0.577753
2.1 1.807699 2.195618 2.221909 1.949309 2.128516 1.367815
2.3 2.03947 2.237211 2.305633 2.15875 2.161046 2.408816
2.5 2.154334 2.282122 2.311104 2.175264 2.175324 2.487444
2.7 2.318824 2.364098 2.390271 2.232623 2.292079 2.602326
2.9 2.374509 2.464817 2.526811 2.246321 2.332821 2.626207
3.1 2.468744 2.527438 2.623165 2.261637 2.335784 2.627086
3.3 2.529209 2.562386 2.641661 2.30133 2.361684 2.627868
3.5 2.568626 2.570972 2.674099 2.303361 2.374183 2.664065
3.7 2.590673 2.652589 2.676686 2.363675 2.386289 2.755612
3.9 2.711469 2.662441 2.701554 2.375425 2.410978 2.779426
4.1 2.745753 2.709625 2.722828 2.399415 2.421417 2.794549
4.3 2.7703 2.722492 2.724241 2.402557 2.461279 2.811444
4.5 2.808867 2.742046 2.735318 2.535892 2.467631 2.825221
4.7 2.850825 2.76393 2.769084 2.537501 2.504708 2.830153
4.9 2.884434 2.807124 2.787544 2.584822 2.523186 2.857571
Table 3
LISS-III entropy of various land cover classes from FCM classification output.
Value of weighting Agriculture bright forest dense forest Agriculture dry Agriculture moist Water Body
exponent ‘m’ land without crop land without crop
1.1 0.005647 0.005647 0 0.005647 0.005647 0
1.3 0.005647 0.005647 0.005647 0.005647 0.005647 0
1.5 0.779405 0.005647 0.005647 0.005647 0.438642 0.005647
1.7 0.931645 0.005647 0.005647 0.005647 0.763175 0.005647
1.9 1.127571 0.391921 0.005647 0.005647 0.763175 0.013378
2.1 1.232044 0.511249 0.563636 0.438642 0.809914 0.024543
2.3 1.32687 0.542586 0.563636 0.791327 0.941826 0.034452
2.5 1.65215 0.569752 0.72791 0.807417 0.941826 0.062745
2.7 1.688436 0.655321 0.969694 0.873741 1.221659 0.241127
2.9 1.754213 0.676722 1.039694 0.937978 1.301497 0.328685
3.1 1.757718 0.847891 1.279952 0.937978 1.377741 0.648482
3.3 1.778683 1.200501 1.282442 1.178673 1.475438 0.735558
3.5 1.819449 1.270168 1.313101 1.390446 1.605331 0.763175
3.7 1.922052 1.447138 1.388364 2.262668 1.721545 0.842982
3.9 1.929977 1.921122 1.460901 2.41688 1.728667 0.996078
4.1 2.03713 1.922168 1.579467 2.431606 1.759878 1.221659
4.3 2.105483 2.009728 1.873025 2.462514 1.852669 1.721545
4.5 2.17118 2.010089 1.878813 2.514286 1.93608 1.852669
4.7 2.195477 2.019947 1.889476 2.555636 1.978665 2.431606
4.9 2.267225 2.08969 1.930662 2.566581 1.996702 2.514286
R. Sharma et al. / Egypt. J. Remote Sensing Space Sci. 23 (2020) 49–55 51
Table 4
LISS-III entropy of various land cover classes from PCM classification output.
Value of weighting Agriculture bright forest dense forest Agriculture dry Agriculture moist Water Body
exponent ‘m’ land without crop land without crop
1.1 0.082956 0.312832 0.145562 0.261924 0.005647 0.060872
1.3 0.082956 0.318478 0.191696 0.261924 0.005647 0.060872
1.5 0.129719 0.318478 0.347175 0.261924 0.005647 0.060872
1.7 0.164975 0.318478 0.347175 0.261924 0.005647 0.060872
1.9 0.280546 0.318478 0.347175 0.261924 0.005647 0.060872
2.1 0.291772 0.318478 0.347175 0.261924 0.005647 0.060872
2.3 0.291772 0.318478 0.347175 0.261924 0.005647 0.066268
2.5 0.291772 0.318478 0.347175 0.261924 0.011271 0.082318
2.7 0.291772 0.318478 0.347175 0.45811 0.011271 0.144367
2.9 0.291772 0.411208 0.347175 0.45811 0.011271 0.169177
3.1 0.291772 0.411208 0.347175 0.45811 0.011271 0.183759
3.3 0.291772 0.411208 0.476853 0.45811 0.011271 0.19811
3.5 0.653291 0.411208 0.476853 0.45811 0.011271 0.221505
3.7 0.768366 0.411208 0.476853 0.511747 0.011271 0.266277
3.9 0.768366 0.422434 0.529408 0.511747 0.011271 0.266277
4.1 0.768366 0.422434 0.529408 0.511747 0.011271 0.362017
4.3 0.768366 0.568146 0.529408 0.511747 0.033543 0.369201
4.5 0.768366 0.568146 0.529408 0.511747 0.033543 0.383181
4.7 0.768366 0.568146 0.529408 0.511747 0.033543 0.433711
4.9 0.768366 0.568146 0.529408 0.511747 0.033543 0.491744
Table 5
LISS-IV entropy of various land cover classes from FCM classification output.
Value of weighting Agriculture bright forest dense forest Agriculture dry Agriculture moist Water Body
exponent ‘m’ land without crop land without crop
1.1 0.005647 0.005647 0 0.005647 0 0.005647
1.3 0.005647 0.005647 0 0.005647 0 0.005647
1.5 0.833044 0.881731 0.005647 0.005647 0.334012 1.040795
1.7 0.93792 0.935478 0.90618 0.90618 0.455002 1.090845
1.9 0.996273 0.987775 1.038942 1.038942 0.493181 1.162071
2.1 1.023482 1.020236 1.094468 1.094468 0.532189 1.171888
2.3 1.060396 1.082032 1.171339 1.171339 0.534927 1.185138
2.5 1.072108 1.14259 1.179385 1.179385 0.568189 1.202948
2.7 1.160526 1.150218 1.181608 1.181608 0.578827 1.234024
2.9 1.165481 1.150845 1.191133 1.191133 0.77937 1.240439
3.1 1.178855 1.185337 1.212169 1.212169 0.883868 1.250455
3.3 1.192471 1.196408 1.226577 1.226577 0.947835 1.361682
3.5 1.215419 1.20713 1.244597 1.244597 0.968205 1.362615
3.7 1.244101 1.216975 1.255464 1.255464 1.166506 1.38732
3.9 1.322855 1.307276 1.265672 1.265672 1.195226 1.400067
4.1 1.341925 1.405059 1.269796 1.269796 1.271948 1.413906
4.3 1.352694 1.44459 1.313847 1.313847 1.271948 1.429775
4.5 1.380341 1.476362 1.338663 1.338663 1.351788 1.437714
4.7 1.417581 1.670922 1.371591 1.371591 1.432229 1.465357
4.9 1.431856 1.687325 1.436047 1.436047 1.456061 2.045423
Table 6
LISS-IV entropy of various land cover classes from PCM classification output.
Value of weighting Agriculture bright forest dense forest Agriculture dry Agriculture moist Water Body
exponent ‘m’ land without crop land without crop
1.1 0.85321 0.720505 0.888267 0.260393 1.047761 0
1.3 1.323762 0.883429 0.938212 0.421313 1.065395 0
1.5 1.333846 0.897783 0.941301 0.425412 1.074278 0.94433
1.7 1.421579 0.921517 0.967753 0.469295 1.079864 0.96242
1.9 1.436194 0.983435 0.969124 0.505332 1.136726 1.073354
2.1 1.45772 1.016671 0.990967 0.532614 1.251742 1.115397
2.3 1.470837 1.027519 1.025035 0.538193 1.263018 1.245537
2.5 1.48336 1.033031 1.025814 0.54505 1.347871 1.343562
2.7 1.598708 1.044512 1.111484 0.587383 1.356683 1.370217
2.9 1.607408 1.060002 1.200183 0.654696 1.362125 1.370722
3.1 1.610917 1.070653 1.20429 0.675792 1.371783 1.382169
3.3 1.612705 1.121911 1.250014 0.719147 1.377224 1.41307
3.5 1.634205 1.292061 1.256776 0.86282 1.380859 1.413125
3.7 1.673346 1.324348 1.323268 0.903478 1.394449 1.491635
3.9 1.713578 1.338266 1.323954 1.349272 1.397588 1.593333
4.1 1.767147 1.488573 1.515676 1.477019 1.397771 1.599144
4.3 1.830625 1.614532 1.589824 1.570496 1.405531 1.688358
4.5 1.839581 1.668217 1.684392 2.452745 1.436348 1.719099
4.7 1.845379 1.893284 1.717801 2.487226 1.438385 1.798186
4.9 1.898212 1.977909 1.743392 2.994849 1.464893 1.882124
52 R. Sharma et al. / Egypt. J. Remote Sensing Space Sci. 23 (2020) 49–55
Table 7
Entropy variation for FCM Classifier.
Table 8
Entropy variation for PCM Classifier.
problems, and the software package developed consists of fuzzy, Ghosh (Kumar et al., 2007) have focused primarily on Support
entropy and based soft classification approaches (Sharma et al., Vector Machine (SVM) and statistical based learning algorithms.
2016, 2017) and has been named as Fuzzy Based classifiers None of these have provided the hybridized model of classification
incorporating Entropy and Noise (FBICET). In general, none of the to incorporate context, entropy, and noise. Further, this tool
commercially available software has incorporated entropy and provides an automated image to image accuracy assessment
contextual based hybridization and SCM based approach to assess approach using entropy.
the accuracy of a classified image. Further, such software packages
provide an option for entropy for multi-spectral remote sensing
3. Programming environment
data at sub-pixel classification. Thus, in this study, it was necessary
to develop a package having the sub-pixel classification algorithms
The coding of Fuzzy Based Image Classification Entropy based
used for different experiments.
package has been done in Java environment. Two types of pro-
We can use FCM (Bezdek et al., 1984) and PCM soft classifiers
grams can be created: applications and applets by Java. The output
(Krishnapuram and Keller, 1993) (Krishnapuram and Keller,
of a Java compiler is not an executable code; rather it is called a
1996) in both modes i.e., unsupervised and supervised, but in this
byte-code. There is an interpreter in java runtime system which
research these classifiers are applied in supervised mode only. This
is used to convert the program into byte code. It is simply a highly
paper is paying attention on the study of an implementation of
capable means of programming a program into byte code. It is
FCM (Fuzzy c-mean) with the tool. This tool plays a vital role in
much easier to permit Java programs to run in a wide selection
term of classification and accuracy assessment approach.
of environments. Therefore, all class objects are dynamically allo-
cated. When there are no references to an object, then the object
2. The requirement of tool is considered as dormant. This tool is platform independent, due
to the JAVA environment.
In present arena of digital image processing software, various
kinds of readymade tools are available to perform digital image
4. Study area and data set
analysis. Since last two decades, a large number of image
processing software has been developed by various commercial
Sitarganj Tehsil (Udham Singh Nagar District, Uttarakhand) area
companies offering a suite of modules related to data input, visual-
is selected for this study. The entropy and fraction images play vital
ization, image enhancement, transformation, classification and
role in term of accuracy assessment and classification.
assessment of accuracy. Some of the leading image processing soft-
ResourceSat-1 (IRS-P6), satellite gives distinctive information and
ware is ERDAS IMAGINE, IDRISI, ENVI, ER MAPPER and e-Cognition.
multi-spectral data at special spatial resolution, while preserving
Similarly, some individual based soft classification have been
the spectral information. AWiFS, LISS-III and LISS-IV of
developed, such as Soft Classification Methods and Accuracy
Resourcesat-1 provide data at spatial resolution of 60 m, 20 m
Assessment Package (SCMAP) developed by Aziz (2004) and Sub-
and 5 m respectively, and they have been used for classification
pixel Multispectral Image Classifier (SMIC) developed by Kumar
or reference fraction images generation in this study shown in
and Gosh (Kumar et al., 2007), which facilitate the image classifica-
Fig. 1.
tion based upon statistical learning algorithms and assessment of
accuracy using FERM only. However, these tools are not able to ful-
fill the objectives defined for this study. The need of tool: Fuzzy 5. Different modules of the package
Based Image Classifier incorporating Entropy and Noise has
emerged, in order to incorporate the entropy and noise in the pro- The software package has five basic modules with main package
cess of satellite image classification. Aziz (2004) incorporated the mentioned as follows, and shown in Fig. 2;
Neural Network based classifier in the study, wherein Kumar and Classification approaches used in this study.
R. Sharma et al. / Egypt. J. Remote Sensing Space Sci. 23 (2020) 49–55 53
1) File Module
2) Display Module
3) Signature Files Module
4) Classifiers Module
5) Help Module
Start
Fig. 6. The GUI for various options available for collecting pure signature data. Stop
the ten sub-pixel classifiers as provided in FBICET package shown Fig. 7. Flow chart for FCM classification.
in Fig. 6. The signature option, allows the user to collect pure ref-
erence pixels so that it can be used as an input by any of the
sub-pixel classifiers provided in this package.
Three norms, namely; Euclidean, Diagonal and Mahalonobis
norms, each induced by specific weight matrix have been incorpo-
rated in D-Matrix, V-C Matrix and DVC Matrix, respectively. How-
ever, in this study only D-Matrix with value 1 has been used for the
classification of satellite imagery.
6. Methodology
absolute indicator for uncertainty and does not require any refer- of the tool is appropriate, however it needs to be tested for other
ence data, there are numerous methods accessible to access uncer- classes also before drawing a concrete conclusion.
tainty of soft classify n for my study. Accuracy measure process
takes account only these samples to evaluate accuracy parameters.
Accuracy assessment process for soft classification required References
classified fraction images and reference classified fraction images
(if available), perform sampling over classified data, apply accuracy Aziz, M.A., 2004. Evaluation of soft classifiers for remote sensing data, unpublished
Ph.D thesis. Indian Institute of Technology Roorkee, Roorkee, India.
assessment method on sampled data and finally produces accuracy
Bezdek, J.C., Ehrlich, R., Full, W., 1984. FCM: The fuzzy c- meansclustering algorithm.
parameters Fig. 9. Comput. Geosci. 10, 191–203.
Bora, D.J., Gupta A.K., 2014. A Comparative study Between Fuzzy Clustering
Algorithm and Hard Clustering Algorithm. Int. J. Computer Trends Technol.
7. Results, analysis and disscussion (IJCTT) 10(2), ISSN: 2231-2803, 108–113.
Dwivedi, R., Kumar, A., Ghosh, S.K., Roy, P., 2012. Optimization of Fuzzy Based Soft
7.1. General working of FCM and PCM with entropy classifier in FBICET Classifiers for Remote Sensing Data. ISP International Archives of the
Photogrammetric, Remote Sensing and Spatial Information Sciences 1, 385–
package 390.
Dwivedi, R.K., Ghosh, S.K., Kumar, A., 2012. Study of Fuzzy based Classifier
7.1.1. Accuracy assessment using entropy Parameter across Spatial Resolution. Int. J. Comput. Appl. (0975–8887) 50(11),
17–24
Uncertainly, the accuracy can be visualized by Entropy. It is an Dwivedi, R.K., Ghosh, S.K., Kumar, A., 2012. Investigation of Image Classification
absolute indicator of uncertainty for image classification i.e. when Techniques For Performance Enhancement. Viewpoint. Int. J. Manage. Technol.
classes is evenly proportioned within pixel then higher entropy ISSN-2229-3825. 3(1), 3–10
Foody, G.M., 1996. Approaches for the production and evaluation of fuzzy land
occur, that shows higher uncertainty and vice versa (Dehghan
cover classifications from remotely-sensed data. Int. J. Remote Sensing 17 (7),
and Ghassemian, 2006). This approach is capable to summarize 1317–1340.
the classification of uncertainty in a single. The entropy may be Jabari, S., Zhang, Y., 2013. Very High Resolution Satellite Image Classification Using
computed from (Zhang and Foody, 2001), Fuzzy Rule-Based Systems. Algorithms 6, 762–781. https://doi.org/10.3390/
a6040762, ISSN 1999-4893, Pg Np. 763-781.
X
C Krishnapuram, R., Keller, J.M., 1993. A possiblistic approach to clustering. IEEE
H¼ lij log2 lij ð1Þ Trans. Fuzzy Syst. 1, 98–108.
Krishnapuram, R., Keller, J.M., 1996. The possibilistic c-means algorithm: Insights
i¼1
andrecommendations. IEEE Trans. Fuzzy Syst. 4 (3), 385–393.
where c is the total no of classes and lij is the estimated member- Kalra, K., Goswami, A.K., Gupta, R., 2013. A comparative study of supervised image
Classification algorithms for satellite images. Int. J. Electrical Electronics Data
ship function of class c for pixel x. The entropy of each pixel is Commun. ISSN: 2320-2084, 1(10)
summed over all pixels to compute the average entropy of the Kumar, A., Ghosh, S.K., Dadhwal, V.K., 2007. Full fuzzy land cover mapping using
whole classification. Constraint is that class memberships of a pixel remote sensing data based on fuzzy c-means and density estimation. Can. J.
Remote sensing 33 (2), 81–87.
across the classes should sum to one. Dehghan, H., Ghassemian, H., 2006. Measurement of uncertainty by the entropy:
application to the classification of MSS data. Int. J. Remote Sensing 27 (18),
4005–4014.
8. Conclusion
Ranjana Sharma, Achal Kumar Goyal, Dwivedi R.K., 2016. A Review of Soft
Classification Approaches on Satellite Image and Accuracy Assessment. In:
In this paper, it is defined to use classifiers FCM and Euclidean Proceedings of Fifth International Conference on Soft Computing for Problem
norm for generating fraction result. This outcome is generated Solving, Springer singapore, vol. 437, pp. 629–639.
Ranjana sharma, Dwivedi, R.K., Garg, P.K., 2017. Performance analysis of FCM and
from LISS-III image of IRS-P6 data. We already know that the clas- PCM soft classifiers for Satellite data. In: Proceedings of the SMART–2017, IEEE
sification method is complete with the assessment of accuracy so Conference ID: 42281, 6th International Conference on System Modeling &
Entropy has been used as the assessment of accuracy for various Advancement in Research Trends, 29th–30th December, 2017, pp. 133–138.
Saha, I., Maulik, U., Bandyopadhyay, S., Plewczynsk, D., 2010. Unsupervised and
land cover classes, i.e., agriculture land, dense forest, water bodies, Supervised learning approaches together for Microarray Analysis. Fundamental
agriculture moist land without crop, and agriculture dry land Informatica 106 (1), 45–73.
without crop. The objective of this research is to investigate, the Tamouk, J., Lotfi, N., Farmanbar, M., 2013. Satellite Image Classification Methods
And Landsat 5tm Bands. Comput. Vis. Pattern Recogn.
amount of accuracy or uncertainty create for spatial data. This Stein, A., 2002. Pattern recognition. In: El-Shaarawi, A.H., Piegorsch, W. (Eds.)
paper proposed the estimating model of spatial data. In brief, this Encyclopedia of environmentrics Vol. 4. Chichester: Wiley & Sons, ISBN: 0-471-
research, on spatial variation has analyzed the total uncertainty 89997-6. 1534-1542.
Tso, B., Mather, P.M., 2001. Classification Methods for Remotely Sensed Data. Taylor
which is not exceeding the reference value. In such cases, entropy
and Francis Inc, New York.
can be used as an absolute measure of an uncertainty (Dehghan Wojtaszek, M.V., Balázsik, V., Jancsó, T., Gulyás, M.H., Meng, Q., 2015. Comparison of
and Ghassemian, 2006). This indicates that the information of three image classification Methods in Urban enviornment. J. Geosci. Environ.
Protection 3, 54–59.
uncertainty is not exceeding more than 3% (the computed entropy
Yannis, S.A., Stefanos, D.K., 1999.-Fuzzy. Fuzzy Image Classification Using Multi-
result has been given in table 1,3,5,7 for FCM and 2,4,6,8 for PCM resolution Neural with Applications to Remote Sensing. Electrical Engineering
classifiers. Measuring the spatial statistics of a satellite image Department, National Technical University of Athens, Zographou 15773, Greece.
using entropy, for six land cover classes can be obtained by using Zhang, J., Foody, G.M., 2001. A fuzzy classification of sub-urban land cover from
remotely sensed imagery. Int. J. Remote Sensing 19, 2721–2738.
Eq (1) i. e. 6*(1/6*log21/6) = 2.585 (Stein, 2002). So the working