=Paper=
{{Paper
|id=Vol-3702/paper24
|storemode=property
|title=Enhancement of Land Cover Classification by Geospatial Data Cube Optimization
|pdfUrl=https://ceur-ws.org/Vol-3702/paper24.pdf
|volume=Vol-3702
|authors=Artem Andreiev,Anna Kozlova,Leonid Artiushyn,Peter Sedlacek
|dblpUrl=https://dblp.org/rec/conf/cmis/AndreievKAS24
}}
==Enhancement of Land Cover Classification by Geospatial Data Cube Optimization==
https://ceur-ws.org/Vol-3702/paper24.pdf
Enhancement of Land Cover Classification by Geospatial
Data Cube Optimization
Artem Andreiev1, Anna Kozlova1, Leonid Artiushyn2 and Peter Sedlacek3
1 Scientific Centre for Aerospace Research of the Earth, Institute of Geological Sciences, National Academy of Sciences of
Ukraine, Olesia Honchara str., 55-b, Kyiv, 01054, Ukraine
2 State Research Institute of Aviation, Hryhoriia Andriuschenka Str., 6-V, Kyiv, 01135, Ukraine
3 University of Zilina, Univerzitna, 8215/1, Zilina, 01026, Slovakia
Abstract
This paper presents the optimization technique to reduce the geospatial data cube size and enhance the
land cover classification. The technique is based on training sample separability. Accordingly, the
Separability Index of the Training Sample (SITS) was developed and used as the object function for the
optimization. In order to test the effectiveness of the optimization technique, the experiment was
conducted. It implied the land cover classification of the highly heterogeneous natural landscapes in the
case of the Shatsky National Natural Park, where the prevailing landscape is wetlands. After the
optimization of the input geospatial data cube, classification enhancement was evidenced by increasing
indicators such as overall accuracy by 0.04 from 0.9 to 0.94 and the kappa coefficient by 0.06 from 0.86
to 0.92. In addition, the data cube size was reduced by 5.55 times from 222 to 40 layers
Keywords
Remote sensing, land cover classification, supervised classification, training sample separability,
geospatial data cube, data optimization1
1. Introduction
Land cover classification is a critical process in remote sensing, providing spatially explicit
information at different scales for numerous environmental applications [1]. Such information is
widely applied to issues that require practical geospatial solutions like land cover change
detection [2], environmental monitoring [3,4], fossil fuel exploration [5], and landmine detection
[6]. Land classification techniques, likewise, play a crucial role in the integration of Earth
observation data into comprehensive interdisciplinary issues on sustainable development goals
achieving [7, 8], in particular, combat climate change and its impacts [9], reverse land degradation
[10] and halt biodiversity loss [11], protect water-related ecosystem for safety water supply [12],
and provide support for food security and sustainable agriculture [13, 14].
Today, most classification methods are mainly divided into supervised and unsupervised [15].
However, in remote sensing, the supervised classification methods are the most appropriate for
the majority of the thematic tasks because applying these methods can establish the
characteristics of the output classes, unlike unsupervised ones. A training sample set is used to
set the characteristics of classes in supervised classification methods. Such a set contains the
signatures of features of each class.
The input data for classification is heterogeneous geospatial data, which can be represented
in the form of raster layers. To combine such layers into a single array, it is customary to form a
geospatial data cube [16]. From ordinary datasets, data cubes differ by integrating different data
types into a coherent and interoperable structure [17, 18]. After the cube's formation, the training
sample's signatures must be determined in each layer. Hence, each layer is a feature of the
training sample.
CMIS-2024: Seventh International Workshop on Computer Modeling and Intelligent Systems, May 3, 2024,
Zaporizhzhia, Ukraine
artem.a.andreev@gmail.com (A. Andreiev); ak.koann@gmail.com (A. Kozlova); artleonid2017@gmail.com (L.
Artiushyn); peter.sedlacek@fri.uniza.sk (P.Sedlacek)
0000-0002-6485-449X (A. Andreiev); 0000-0001-5336-237X (A. Kozlova); 0000-0002-7488-7244 (L. Artiushyn);
0000-0002-7481-6905 (P.Sedlacek)
ยฉ 2024 Copyright for this paper by its authors.
Use permitted under Creative Commons License Attribution 4.0 International (CC BY 4.0).
CEUR
ceur-ws.org
Workshop ISSN 1613-0073
Proceedings
A geospatial data cube for specific classification tasks can comprise numerous layers.
Including multitemporal data, a data cube aims to distinguish dynamic objects that change
significantly during a specific period or vary much from each other at different stages of their
development, e.g. vegetation cover, wetlands in particular [12]. Different physical aspects of
multisource data, e.g. optical and radar, highlight diverse object traits and variations in land cover
types [19]. Multiple ancillary data, e.g. data on geomorphology, hydrology, or phenology, help to
differentiate land cover types due to their context [20].
However, the redundancy of the geospatial data cube causes two significant problems [21].
Firstly, the processing of such a data cube has high computational complexity. Secondly, since the
signatures of the training sample are defined in each layer of the geospatial data cube, the
separability of the training sample can be reduced if the layers are either incorrectly created or
irrelevant to the selected thematic task. In turn, the low separability of the training sample leads
to a decrease in classification accuracy [22].
In light of the above, optimization of the geospatial data cube is seen as a solution to the
mentioned problems [23, 24]. Among approaches to reduce input data, the Principal Component
Analysis (PCA) [25] and the Minimum Noise Fraction (MNF) [26] are the most widely used. There
are also similar methods, for example, Noise-Adjusted Principal Components (NAPC) [27],
Independent Component Analysis (ICA) [28], Non-Negative Matrix Factorization (NMF) [29] and
Spatio-Spectral Decomposition (SSD) [30]. However, a common disadvantage of the considered
approaches is that they do not consider the training sample's structure (in particular
separability) and the selected classifier's specificity.
The presented study aims to enhance land cover classification by selecting the cube layers, the
training sample separability of which will be the highest among other options. For this purpose,
the optimization technique of the geospatial data cube was developed. It has two goals:
enhancement of land cover classification and reduction of geospatial data cube size.
Hence, in the relevant sections of this article, the separability assessment of the training
sample, the optimization technique of the geospatial data cube, and the experiment conducted to
demonstrate the effectiveness of the developed technique are described.
2. Methods
This section presents the optimization technique of a geospatial data cube. Since this optimization
is based on training sample separability, the objection function is the developed separability
index of the training sample (SITS). Thus, the training sample separability assessment is also
presented below as an algorithm for SITS calculation.
2.1. Assessment of the training sample separability
Separability is one of the training sample characteristics that affect classification accuracy.
This characteristic shows the extent to which signatures representing different classes do not
overlap. A low degree of separability is inherent in a high level of training sample mixing. In turn,
this leads to a significant number of misclassified objects in the classification. Thus, the training
sample separability is directly proportional to the classification accuracy.
The algorithm depicted in the flowchart(Figure 1) describes the separability assessment of the
training sample.
Figure 1: Algorithm of the separability assessment of the training sample
The first step implies classifier training by the training sample. Importantly, the supervised
classification method must be the same as the one selected to classify the geospatial data cube
further. Moreover, due to the proposed, separability depends on its structure (i.e., the set of
layers) and the selected supervised classification method.
In the second step, the classifier is used to classify each signature from the training sample set.
The third step is the formation of the confusion matrix [31] for the classification obtained in
the previous step.
The fourth and final step is calculating the SITS. This index quantifies the separability of
training samples by measuring the ratio of correctly classified training samples to the total
number of training samples. In other words, SITS equals the overall accuracy [31] based on the
confusion matrix obtained in the previous step. The calculation of the SITS is shown in the
following formula:
โ๐พ ๐=1 ๐ฅ๐๐ (1)
๐๐ผ๐๐ = ,
๐
where ๐พ is the number of classes, N is the total number of training sample signatures, ๐ฅ๐๐ is the
number of class i signatures classified as class i (i.e. diagonal elements of the obtained confusion
matrix that correspond to correctly classified signatures).
The values of the considered index range from 0 to 1. In this case, the value 0 shows that the
training sample is entirely mixed (minimum separability), and the value 1 corresponds to the
training sample, which is entirely separable (maximum separability).
2.2. Optimization technique
This technique is an optimization procedure that aims to reduce the number of layers of the
geospatial data cube and increase the separability of the training sample, the signatures of which
are defined in each layer of this cube. The objective function implies using the SITS. Thus,
geospatial data cube optimization can be described as a search of the minor number of cube layers
for which the training sample has the highest SITS value among all other sets of cube layers.
The flowchart of the technique algorithm is shown in Figure 2.
Figure 2: Algorithm of the optimization technique
The initial data and their characteristics will be introduced below for a detailed description of
the technique algorithm.
Let the initial geospatial data cube have the following form:
๐บ๐ถ๐ท๐๐๐๐ก๐๐๐ = {๐ฟ1 , ๐ฟ2, โฆ , ๐ฟ๐ },
where ๐ฟ๐ is the layer i of the geospatial data cube, and N is the total number of layers included
in the initial geospatial data cube.
Then, as shown in Figure 2, the SITS value of the initial training sample, which has the
signatures defined in each layer of the initial geospatial data cube, is first calculated. Let this value
be ๐๐ผ๐๐๐๐๐๐ก๐๐๐ .
Let us assign the values obtained above to the variables corresponding to the optimal set of
layers of the geospatial data cube and the corresponding value of the SITS. Thus, we have:
๐บ๐ถ๐ท๐๐๐ก๐๐๐๐ โ ๐บ๐ถ๐ท๐๐๐๐ก๐๐๐ ,
๐๐ผ๐๐๐๐๐ก๐๐๐๐ โ ๐๐ผ๐๐๐๐๐๐ก๐๐๐ .
Next, an iterative procedure follows, in which the following steps are performed at each
iteration.
Step 1. At the iteration i, the current geospatial data cube ๐บ๐ถ๐ท๐๐๐ก๐๐๐๐ , consisting of N-(i-1)
layers, is decomposed into N-(i-1) cubes. Each newly formed cube is obtained by discarding one
of the layers from the current cube. Then, each of the newly created geospatial data cubes will
have the following form:
๐บ๐ถ๐ท1โฒ = ๐บ๐ถ๐ท โ {๐ฟ1 } = {๐ฟ2 , ๐ฟ3, โฆ , ๐ฟ๐โ(๐โ1) },
๐บ๐ถ๐ท2โฒ = ๐บ๐ถ๐ท โ {๐ฟ2 } = {๐ฟ1 , ๐ฟ3 โฆ , ๐ฟ๐โ(๐โ1) },
โฆ
โฒ
๐บ๐ถ๐ท๐โ(๐โ1) = ๐บ๐ถ๐ท โ {๐ฟ๐โ(๐โ1) } = {๐ฟ1 , โฆ , ๐ฟ๐โ(๐โ1)โ1 }.
Since the number of layers decreases by one at each iteration, the obtained cubes will contain
N-i layers. Therefore, the following is valid:
โฒ
|๐บ๐ถ๐ท1โฒ | = |๐บ๐ถ๐ท2โฒ | = โฏ = |๐บ๐ถ๐ท๐โ(๐โ1) | = ๐ โ ๐.
Thus, the generated cubes can be written in the form of the following set:
โฒ
๐ถ = {๐บ๐ถ๐ท1โฒ , ๐บ๐ถ๐ท2โฒ , โฆ , ๐บ๐ถ๐ท๐โ(๐โ1) }.
Step 2. For each newly formed cube, the SITS value is calculated for the training sample, the
signatures of which are defined in each cube layer. Then, the value of SITS for a particular cube
๐บ๐ถ๐ท๐กโฒ will be denoted as ๐๐ผ๐๐๐กโฒ . Thus, a set containing the value of the SITS for each newly formed
cube will be obtained:
โฒ
๐ = {๐๐ผ๐๐1,โฒ ๐๐ผ๐๐2โฒ , โฆ , ๐๐ผ๐๐๐โ(๐โ1) }.
Step 3. Among the obtained cubes, the one with the highest value of the SITS is selected. Such
a cube will be denoted as ๐บ๐ถ๐ท๐ . The selected cube can be expressed as follows:
๐บ๐ถ๐ท๐ = {๐บ๐ถ๐ท๐กโฒ โ ๐ถ|๐๐ผ๐๐๐กโฒ = max{๐}}.
๐๐ผ๐๐๐ denotes the value of the SITS of the geospatial data cube ๐บ๐ถ๐ท๐ .
Step 4. The values of the variables ๐๐ผ๐๐๐ and ๐๐ผ๐๐๐๐๐ก๐๐๐๐ are compared, and two options are
considered:
1) if ๐๐ผ๐๐๐ < ๐๐ผ๐๐๐๐๐ก๐๐๐๐ , then the execution of the optimization algorithm will be interrupted,
and further steps will be ignored. The optimal geospatial data cube will be the one obtained in
the previous iteration, namely ๐บ๐ถ๐ท๐๐๐ก๐๐๐๐ . Accordingly, the SITS of the training sample of the
corresponding cube has the value ๐๐ผ๐๐๐๐๐ก๐๐๐๐ .
2) if ๐๐ผ๐๐๐ โฅ ๐๐ผ๐๐๐๐๐ก๐๐๐๐ , then the variable ๐๐ผ๐๐๐๐๐ก๐๐๐๐ is assigned the value of the variable
๐๐ผ๐๐๐ , i.e.:
๐๐ผ๐๐๐ โ ๐๐ผ๐๐๐๐๐ก๐๐๐๐ .
Step 5. At this step, as at the previous one, two options are considered:
1) if the number of layers of the obtained cube ๐บ๐ถ๐ท๐๐๐ก๐๐๐๐ is 1, i.e.:
|๐บ๐ถ๐ท๐๐๐ก๐๐๐๐ | = 1,
then the execution of the optimization algorithm will be interrupted. The optimal geospatial data
cube will be the one obtained at the current iteration, namely โ ๐บ๐ถ๐ท๐๐๐ก๐๐๐๐ . Accordingly, the SITS
value of the training sample of the corresponding cube is โ ๐๐ผ๐๐๐๐๐ก๐๐๐๐ .
2) if the number of layers of the obtained geospatial data cube ๐ฎ๐ช๐ซ๐๐๐๐๐๐๐ is greater than 1,
i.e.:
|๐บ๐ถ๐ท๐๐๐ก๐๐๐๐ | > 1,
then a new iteration will start, and the actions specified in step 1 will be performed again.
The result of the described optimization procedure is the geospatial data cube with the set of
layers that achieves the highest value of the SITS among all other considered sets. This geospatial
data cube will be used for further land cover classification.
3. Experiment
The experiment was conducted to test the effectiveness of the developed technique. It consisted
of carrying out a classification of the selected study area. For this purpose, an initial geospatial
data cube with an excessive number of layers was formed. Then, the optimization technique was
applied, resulting in an optimized cube. Finally, two classifications were obtained โ before and
after the optimization.
3.1. Study area
Since 2007, the Ukrainian network of test sites has provided validation and calibration of
various remote sensing techniques and satellite-based products, including land cover
classification [32]. The proposed technique was tested at the site within the Shatsk National
Natural Park (SNNP). It is situated in the northwest of Ukraine, within Volyn' oblast, between 51ยบ
28'25"N and 23ยบ 49'29"E. The SNNP encompasses highly heterogeneous natural landscapes, like
forests, peat bogs, transitional mires, meadows, and lakes.
The site comprises more than 100 georeferenced sample plots and gives comprehensive
ground truth information about the representative landscapes of the West Polissia region (Figure
3).
Figure 3: Location of the study area and sample plots within the Shatsk National Natural Park.
The background is the true-colored composite of the Sentinel-2 Multispectral Instrument (MSI)
image acquired on 1 June 2018
3.2. Training sample
Six broad land cover classes that characterize the study area were defined: artificial surfaces,
tree-covered areas, grassland, agricultural areas, water bodies, and wetlands.
The given classes varied considerably both in spatial extent and heterogeneity. The smallest
class of artificial surfaces included diverse features of built-up areas and transport units. While
the biggest ones, like tree-covered areas and wetlands, included various sub-types that could still
be quite homogenous due to the big extent. Water bodies represent the most homogeneous class.
Therefore, the number of training pixels of each class also varied disproportionally. The overall
number of all training pixels accounted for 6474. Table 1 shows labels, descriptions, and training
pixel amounts for the land cover classes assigned for the experiment.
Table 1
The classification scheme used in the experiment
# Land Cover Class Description Training pixels
Urban public and industrial built-up areas,
1 Artificial surfaces 319
transport units, and construction sites
Broadleaved, coniferous, mixed and swamped
2 Tree-covered areas 2313
forests, orchards, roadside tree lines
Natural herbaceous vegetation, permanent
3 Grasslands 634
grasslands of natural origin, pastures
Arable land, permanent crops, fallow lands,
4 Agricultural areas 887
heterogeneous agricultural areas, open soils
Lakes, rivers and streams of natural origin,
5 Water bodies 634
including man-made reservoirs and canals
Non-forested areas of peat bogs, transitional
6 Wetlands 1687
mires, eutrophic marshes, and reed beds
3.3. Initial geospatial data cube
The experimental classification focuses on wetlands, the prevailing landscape of the test site
and one of the most important for conservation within the SNNP. Evident differences in the
seasonal development of wetlands and other vegetative land cover classes help distinguish them
and require the application of multitemporal data [12].
The primary data source for forming the geospatial data cube was Sentinel-2 satellite imagery
[33]. The images were selected for 4 dates (04.07.2018, 05.12.2018, 06.01.2018 and 10.14.2018)
with minimal or no cloudiness.
Each Sentinel-2 image contains 13 spectral bands. At the preprocessing stage, atmospheric
correction was performed for each image to eliminate the influence of the atmosphere and
calculate the pixel values corresponding to the surface reflectance (bottom of atmosphere).
During this procedure, 3 bands (B1, B9 and B10) that consider the effects of aerosols and water
vapour on reflectance were removed. The spectral bands of the Sentinel-2 image have different
spatial resolutions, namely 10, 20 and 60 meters.
Next, a complete set of normalized difference indices was calculated for each image. The
following combnatorial formula is used to calculate such an index:
๐๐ โ ๐๐
๐ผ๐๐๐๐ฅ = , ๐ โ ๐,
๐๐ + ๐๐
where ๐๐ is the spectral band i.
This set could be presented in the following form:
๐๐ โ ๐๐
๐๐ท๐ผ = { ๐ โ {1,2 โฆ ,10}, ๐ โ ๐}.
๐๐ + ๐๐ |๐,
The cardinality of this set (i.e. number of normalized difference indices of one image) is
calculated by the formula below:
๐! 10! 8! โ 9 โ 10 90
|๐๐ท๐ผ| = ๐ถ๐๐ = = = = = 45,
๐! (๐ โ ๐)! 2! (10 โ 2)! 1 โ 2 โ 8! 2
where variable n corresponds to the number of image bands, and m is the number of
arguments in the index calculation formula.
As a result, 45 spectral indices were obtained for each image.
Another component of the input cube was the geomorphological data obtained from the ALOS
PALSAR DEM [34]. In particular, this data contains the height above sea level and the slope. Their
spatial resolution is 12.5 m.
All the above-described data must be spatially regularized to form the input geospatial data
cube. It involves bringing all layers to the same spatial resolution, map projection, and size. With
this in mind, all layers were scaled to a spatial resolution of 10 m, transformed to Universal
Transverse Mercator Projection, Zone 34N (EPSG:32634) and resized so that all layers lie within
the study area.
So, the input geospatial data cube contained 222 raster layers, namely 40 spectral bands of 4
different time Sentinel-2 satellite images, 180 corresponding spectral indices, and 2 raster layers
of geomorphological parameters. The layers of the input cube are described in Table 2.
Table 2
Layers of the initial geospatial data cube
Date Spectral bands Spectral indices Geomorphological
(DD/MM/YYYY) data
07.04.2018 10 45
12.05.2018 10 45
2
01.06.2018 10 45
14.10.2018 10 45
Hence, signatures of the initial training sample were assigned in each layer of the initial cube.
To assess the separability of the training sample, the SITS was calculated using Formula 1:
โ๐พ๐=1 ๐ฅ๐๐ 6469
๐๐ผ๐๐๐ผ๐๐๐ก๐๐๐ = = โ 0.9992.
๐ 6474
As seen above, this training sample had 5 misclassified signatures.
3.4. Optimized geospatial data cube
After applying the developed optimization technique, the geospatial data cube size was
reduced from 222 to 40 layers. The selected layers are listed below in Table 3.
Table 3
Layer of the optimized geospatial data cube
Date Spectral bands Spectral indices Geomorphological
(DD/MM/YYYY) data
07.04.2018 0/10 4/45
12.05.2018 4/10 0/45
1/2
01.06.2018 2/10 17/45
14.10.2018 1/10 11/45
Along with the cube optimization, the training sample signatures were reassigned according
to the selected cube layers. The separability of that training sample was assessed by the SITS
calculated below:
โ๐พ๐=1 ๐ฅ๐๐ 6474
๐๐ผ๐๐๐๐๐ก๐๐๐๐ = = = 1.
๐ 6474
This training sample had no misclassified signatures, so the optimized training sample was
entirely separated.
3.5. Land cover classifications
In order to test the effectiveness of the developed technique, the classifications using the initial
cube and the optimized one were compared.
Firstly, the classification was obtained using the initial cube and the appropriate training
sample. Then, the classification was obtained using the optimized cube and the appropriate
training sample. These classifications are depicted in Figure 4.
a)
b)
Figure 4: Land cover maps of the study area were obtained using a) the initial cube and b) the
optimized cube
The classifications above were obtained using Mahalanobis distance [35] as a supervised
classification method. Exactly for this method, the separability assessment of the training sample
was carried out for both initial and final geospatial data cubes.
3.6. Accuracy assessment
Classification accuracy assessment involved independent verification of initial and final land
cover maps using proportionate stratified random samplings. This sampling technique produces
sample set sizes directly related to the size of the classes and is widely used in assessing the
classification accuracy of classes disproportionate in their extent. As the required sample size for
a class, 0.01% of the total classified pixels of this class were analyzed. Thus, test sample sets were
equal to 300 pixels for each land cover map. Satellite images (QuickBird) of high spatial
resolution, available for 2018 in the Google Earth Pro app, were used for verification as reference
data.
Table 4 shows the confusion matrix of the initial land cover map. In addition to the two
dimensions ("Reference" and "Prediction"), this matrix shows metrics such as producer accuracy
(PA) and user accuracy (UA) for each class [31].
Table 4
Confusion matrix of the initial land cover map
Reference
Total
1 2 3 4 5 6 UA
(pixels)
1.Artificial
14 0 1 2 7 0 24 0,58
surfaces
2.Tree-
covered 0 134 0 0 1 0 135 0,99
areas
Prediction
3.Grasslands 0 3 29 5 0 1 38 0,76
4.Other
0 0 0 15 0 0 15 1
lands
5.Water
0 0 0 0 53 0 53 1
bodies
6.Wetlands 1 0 6 2 1 25 35 0,71
Total
15 137 36 24 62 26 300
(pixels)
PA 0,93 0,98 0,81 0,63 0,85 0,96
The accuracy of the obtained land cover classification was assessed by indicators of overall
accuracy and the kappa coefficient [31]. The overall accuracy value was calculated by the
following formula:
โ๐พ๐=1 ๐ฅ๐๐ 14 + 134 + 29 + 15 + 53 + 25 270
๐๐ด = = = = 0.9,
๐ 300 300
where ๐พ is the number of classes, N is the total number of test samples, ๐ฅ๐๐ is diagonal element
i of the confusion matrix (i.e. number of correctly classified samples of class i). The value of the
kappa coefficient was obtained following the calculations below:
๐ โ โ๐พ ๐พ ๐พ ๐พ
๐=1 ๐ฅ๐๐ โ โ๐=1(โ๐=1 ๐ฅ๐๐ โ โ๐=1 ๐ฅ๐๐ ) 300 โ 270 โ 24779 56221
๐พ๐๐๐๐ = 2 ๐พ ๐พ ๐พ = 2
= โ 0.86.
๐ โ โ๐=1(โ๐=1 ๐ฅ๐๐ โ โ๐=1 ๐ฅ๐๐ ) 300 โ 24779 65221
Table 5 shows the confusion matrix of the final land cover map.
Table 5
Confusion matrix of the final land cover map
Reference
Total
1 2 3 4 5 6 UA
(pixels)
1.Artificial
8 0 0 1 1 0 10 0,8
surfaces
2.Tree-
covered 0 128 0 0 0 0 128 1
areas
Prediction
3.Grasslands 0 6 40 1 0 0 47 0,85
4.Other
1 0 0 14 0 0 15 0,93
lands
5.Water
0 0 0 0 56 0 56 1
bodies
6.Wetlands 1 1 1 5 0 36 44 0,82
Total
10 135 41 21 57 36 300
(pixels)
PA 0,8 0,95 0,98 0,67 0,98 1
The same indicators were selected for the final land cover classification as for the initial one.
Thus, the value of overall accuracy value is stated below:
๐๐ โ๐พ ๐ฅ 8+128+40+14+56+36 282
๐๐ด = ๐=1 ๐
= 300
= 300 = 0.94.
Then, the following calculations are for the kappa coefficient value:
๐ โ โ๐พ ๐พ ๐พ ๐พ
๐=1 ๐ฅ๐๐ โ โ๐=1(โ๐=1 ๐ฅ๐๐ โ โ๐=1 ๐ฅ๐๐ ) 300 โ 282 โ 24779 59821
๐พ๐๐๐๐ = ๐พ ๐พ ๐พ = = โ 0.92.
2
๐ โ โ๐=1(โ๐=1 ๐ฅ๐๐ โ โ๐=1 ๐ฅ๐๐ ) 3002 โ 24779 65221
4. Discussion
The aims of the developed optimization technique are stated as layers reduction of the geospatial
data cube and enhancement of the classification. Thus, the experiment result should be
considered in terms of these two aspects.
Firstly, the size of the optimized cube was 40 layers, whereas the initial one contained 222
layers. Therefore, the number of layers was reduced by 5.55 times.
Secondly, classification enhancement was evidenced by increasing indicators such as overall
accuracy and the kappa coefficient. Namely, the overall accuracy increased by 0.04 from 0.9 to
0.94, and the kappa coefficient increased by 0.06 from 0.86 to 0.92. Since the classification
focused on the wetlands, the accuracy of this class should be considered individually. Thereby,
both user and producer accuracy of wetlands class were significantly increased, viz. by 0.11 from
0.71 to 0.82 and by 0.4 from 0.96 to 1, respectively.
5. Conclusion
This article presents an optimization technique to reduce geospatial data cube size and enhance
land cover classification. The technique is based on the separability of the training sample, which
is defined in each layer of the geospatial data cube. To assess the separability, the appropriate
index (i.e. SITS) was developed and used as an object function in the technique frame. The
algorithm of the optimization technique implies stepwise band discarding to define the optimal
set of the geospatial data cube layers. Such a set has the highest value of SITS among other options.
The conducted experiment implied techniques application to the land cover classification of
the highly heterogeneous natural landscapes in the case of the Shatsky National Natural Park.
This classification covered six land cover classes where wetlands are prevailing. The technique's
effectiveness was approved by geospatial data cube reduction and classification accuracy
enhancement, evidenced by the increase in such indicators as overall accuracy and kappa
coefficient.
Further research should be aimed at technique application in other study areas and thematic
tasks. Also, the separability assessment of the training sample could be extended by additional
criteria. For example, the kappa coefficient could substitute overall accuracy as the basis of the
developed separability index.
References
[1] A.M. Melesse, Q. Weng, P.S. Thenkabail, G.B. Senay, Remote sensing sensors and applications
in environmental resources mapping and modelling, Sensors, vol. 7, no. 12, pp. 3209โ3241
(2007). doi: 10.3390/s7123209
[2] A.H. Chughtai, H.U. Abbasi, ฤฐ. R. Karaล, A review on change detection method and accuracy
assessment for land use land cover, Remote Sensing Applications: Society and Environment,
vol. 22, p. 100482 (2021). doi: 10.1016/j.rsase.2021.100482
[3] E. Zaitseva, S. Stankevich, A. Kozlova, I. Piestova, V. Levashenko, P. Rusnak, Assessment of the
risk of disturbance impact on primeval and managed forests based on earth observation data
using the example of Slovak Eastern Carpathians, IEEE Access, vol. 9, pp. 162847โ162856
(2021). doi: 10.1109/access.2021.3134375
[4] A. Kozlova, S. Stankevich, M. Svideniuk, A. Andreiev, Quantitative Assessment of Forest
Disturbance with C-Band SAR Data for Decision Making Support in Forest Management,
Lecture notes on data engineering and communications technologies, pp. 548โ562 (2021).
doi: 10.1007/978-3-030-82014-5_37
[5] M.A. Popov, ะขopolnytskyi ะ.V., O.V. Titarenko, S.A. Stankevich, A.A. Andreiev, Forecasting gas
and oil potential of subsoil plots via co-analysis of satellite, geological, geophysical and
geochemical information by means of subjective logic, WSEAS Transactions on Computer
Research, vol. 8, pp. 90โ101 (2020). doi: 10.37394/232018.2020.8.11
[6] M.A. Popov, S.A. Stankevich, S.P. Mosov, O.V. Titarenko, S.S. Dugin, S.I. Golubov, A.A. Andreiev,
Method for Minefields Mapping by Imagery from Unmanned Aerial Vehicle, Advances in
Military Technology, vol. 17, no. 2, pp. 211โ229 (2022). doi: 10.3849/aimt.01722
[7] A. Andries, S. Morse, R.J. Murphy, J.M. Lynch, E. Woolliams, J. Fonweban, Translation of Earth
observation data into sustainable development indicators: An analytical framework,
Sustainable Development, vol. 27, no. 3, pp. 366โ376 (2018). doi: 10.1002/sd.1908
[8] G. Scott, A. Rajabifard, Sustainable development and geospatial information: a strategic
framework for integrating a global policy agenda into national geospatial capabilities,
Geospatial Information Science, vol. 20, no. 2, pp. 59โ76 (2017). doi:
10.1080/10095020.2017.1325594
[9] M. Popov, S. Stankevich, Y. Kostyuchenko, A. Kozlova, Analysis of Local Climate Variations
Using Correlation between Satellite Measurements of Methane Emission and Temperature
Trends within Physiographic Regions of Ukraine, International Journal of Mathematical,
Engineering and Management Sciences, vol. 4, no. 2, pp. 276โ288 (2019). doi:
10.33889/ijmems.2019.4.2-023
[10] O. Dubovyk, The role of Remote Sensing in land degradation assessments: opportunities and
challenges, European Journal of Remote Sensing, vol. 50, no. 1, pp. 601โ613 (2017). doi:
10.1080/22797254.2017.1378926
[11] E. Agrillo, F. Filipponi, A. Pezzarossa, L. Casella, D. Smiraglia, A. Orasi, F. Attorre, A. Taramelli,
Earth Observation and Biodiversity Big Data for forest habitat types classification and
mapping, Remote Sensing, vol. 13, no. 7, p. 1231 (2021). doi: 10.3390/rs13071231
[12] I. Dronova, P. Gong, L. Wang, L. Zhong, Mapping dynamic cover types in a large seasonally
flooded wetland using extended principal component analysis and object-based
classification, Remote Sensing of Environment, vol. 158, pp. 193โ206 (2015). doi:
10.1016/j.rse.2014.10.027
[13] P. Defourny, S. Bontemps, N. Bellemans, C. Cara, G. Dedieu, E. Guzzonato, O. Hagolle, J. Inglada,
L. Nicola, T. Rabaute, M. Savinaud, C. Udroiu, S. Valero, A. Bรฉguรฉ, J. Dejoux, A. Harti, J. Ezzahar,
N. Kussul, K. Labbassi, V. Lebourgeois, M. Zhang, T. Newby, A. Nyamugama, N. Salh, A.
Shelestov, V. Simonneaux, P. Traorรฉ, S. Traorรฉ, B. Koetz, Near real-time agriculture
monitoring at national scale at parcel resolution: Performance assessment of the Sen2-Agri
automated system in various cropping systems around the world, Remote Sensing of
Environment, vol. 221, pp. 551โ568 (2019). doi: 10.1016/j.rse.2018.11.007
[14] S. Fritz, I. McCallum, L. You, A. Bun, E. Moltchanova, M. Duerauer, F. Albrecht, C. Schill, C.
Perger, P. Havlรญk, A. Mosnier, P. Thornton, U. WoodโSichra, M. Herrero, I. BeckerโReshef, C.
Justice, M. Hansen, P. Gong, S. Aziz, M. Obersteiner, Mapping global cropland and field size,
Global Change Biology, vol. 21, no. 5, pp. 1980โ1992 (2015). doi: 10.1111/gcb.12838
[15] R. Sathya, A. Abraham, Comparison of supervised and unsupervised learning algorithms for
pattern classification, International Journal of Advanced Research in Artificial Intelligence,
vol. 2, no. 2 (2013). doi: 10.14569/ijarai.2013.020206
[16] D. Montero, G. Kraemer, A. Anghelea, C.A. Camacho, G. Brandt, G. Camps-Valls, F. Cremer, I.
Flik, F. Gans, S. Habershon, C. Ji, T. Kattenborn, L. Martรญnez-Ferrer, F. Martinuzzi, M.
Reinhardt, M. Sรถchting, K. Teber, M. Mahecha, Data Cubes for Earth System research:
Challenges ahead, EarthArXiv (California Digital Library) (2023). doi: 10.31223/x58m2v
[17] M. Schramm, E. Pebesma, M. Milenkoviฤ, L. Foresta, J. Dries, A. Jacob, W. Wagner, M. Mohr, M.
Neteler, M. Kadunc, T. Miksa, P. Kempeneers, J. Verbesselt, B. Gรถรwein, C. Navacchi, S.
Lippens, J. Reiche, The OpeNEO APIโHarmonizing the use of earth observation cloud services
using virtual Data Cube functionalities, Remote Sensing, vol. 13, no. 6, p. 1125 (2021). doi:
10.3390/rs13061125
[18] L.M. Estupiรฑรกn-Suรกrez, F. Gans, A. Brenning, V.H. Gutierrez-Velez, M.C. Londoรฑo, D.E. Pabon-
Moreno, G. Poveda, M. Reichstein, B. Reu, C. Sierra, U. Weber, M.D. Mahecha, A Regional Earth
System Data Lab for Understanding Ecosystem Dynamics: An Example from Tropical South
America, Frontiers in Earth Science, vol. 9 (2021). doi: 10.3389/feart.2021.613395
[19] T. Hermosilla, M.A. Wulder, J.C. White, N.C. Coops, Land cover classification in an era of big
and open data: Optimizing localized implementation and training data selection to improve
mapping outcomes, Remote Sensing of Environment, vol. 268, p. 112780 (2022). doi:
10.1016/j.rse.2021.112780
[20] J.R.M. Flรณrez, I. Lizarazo, Land cover classification at three different levels of detail from
optical and radar Sentinel SAR data: a case study in Cundinamarca (Colombia), Dyna-
colombia, vol. 87, no. 215, pp. 136โ145 (2020). doi: 10.15446/dyna.v87n215.84915
[21] H. Li, J. Cui, X. Zhang, Y. Han, L. Cao, Dimensionality reduction and classification of
hyperspectral remote sensing image feature extraction, Remote Sensing, vol. 14, no. 18, p.
4579 (2022). doi: 10.3390/rs14184579
[22] I. Piestova, A. Kozlova, A. Andreiev, J. Rabcan Local Quality Improvement of Multispectral
Imagery Classification with Radiometric-spatial Feedback. Computer Modeling and
Intelligent Systems 2864:158โ168 (2021). https://doi.org/10.32782/cmis/2864-14
[23] F. Luo, L. Zhang, B. Du, L. Zhang, Dimensionality reduction with enhanced hybrid-graph
discriminant learning for hyperspectral image classification, IEEE Transactions on
Geoscience and Remote Sensing, vol. 58, no. 8, pp. 5336-5353 (2020).
[24] H. Huang, G. Shi, H. He, Y. Duan, F. Luo, Dimensionality reduction of hyperspectral imagery
based on spatialโspectral manifold learning, IEEE transactions on cybernetics, vol. 50, no. 6,
pp. 2604-2616 (2019).
[25] N. Salem, S. Hussein,. Data dimensional reduction and principal components analysis.
Procedia Computer Science, 163, 292โ299 (2019).
https://doi.org/10.1016/j.procs.2019.12.111
[26] A. Green, M. Berman, P. Switzer, M. Craig, A transformation for ordering multispectral data
in terms of image quality with implications for noise removal. IEEE Transactions on
Geoscience and Remote Sensing, 26(1), 65โ74 (1988). https://doi.org/10.1109/36.3001
[27] C. Chang, Q. Du, Interference and noise-adjusted principal components analysis. IEEE
Transactions on Geoscience and Remote Sensing, 37(5), 2387โ2396 (1999).
https://doi.org/10.1109/36.789637
[28] A. Hyvรคrinen, J. Karhunen, E. Oja, Independent Component Analysis. John Wiley & Sons.
(2001).
[29] D. D. Lee, H. S. Seung, Algorithms for Non-negative Matrix Factorization. In Advances in
Neural Information Processing Systems 13 NIPS 2000, 556-562 (2001).
[30] X. Kong, Y. Zhao, J. Chan, J. Xue, Hyperspectral image restoration via Spatial-Spectral residual
total variation regularized Low-Rank tensor decomposition. Remote Sensing, 14(3), 511
(2022). https://doi.org/10.3390/rs14030511
[31] M.O. Popov, Methodology of accuracy assessment of classification of objects on space images,
Journal of Automation and Information Sciences, vol. 39, pp. 1โ10 (2007). doi: 10.1615/J
Automat Inf Scien.v39.i1.50
[32] V.I. Lyalko, M.A. Popov, S.A. Stankevich, J.I. Zelyk, S.V. Cherny, Calibration/Validation Test
Sites in Ukraine: current state and directions of further research and development, Ukrainian
Metrological Journal, vol. 2, pp. 15-26 (2014).
[33] I. Shurmer, F. Marchese, J.-M. Morales-Santiago, P. P. Emanuelli, Sentinels Optical
Communications Payload (OCP) Operations: From Test to In-Flight Experience, Paper
Presentation, 2018 SpaceOps Conference, p. 24. doi: 10.2514/6.2018-2654
[34] J. S. Stewart, Combining satellite data with ancillary data to produce a refined land-use/land-
cover map, pp. 1-10 (1998). doi: 10.3133/wri974203
[35] P. C. Mahalanobis, On the Generalized Distance in Statistics, Proceedings of the National
Institute of Sciences of India, vol. 2, no. 1, pp. 49โ55 (1936).