Investigation of The Similarity Algorithm of The Satellite Images Storage System For Stability On The Basis of Haar Wavelets According To Tikhonov
Investigation of The Similarity Algorithm of The Satellite Images Storage System For Stability On The Basis of Haar Wavelets According To Tikhonov
Investigation of The Similarity Algorithm of The Satellite Images Storage System For Stability On The Basis of Haar Wavelets According To Tikhonov
DOI: https://doi.org/10.3897/arb.v31.e06
Abstract
A model of the automated space monitoring system (ASMS) for the presence of waste
disposal facilities (WDFs) is being developed in this paper. One of the components of the model is the
space imagery storage unit, which allows not only to improve the performance of environmental
authorities, but also to enhance the effectiveness of WGF monitoring (space, aerological, and ground
based). The article also investigates the problem of searching in space image archives of similarities
based on Haar wavelet transformations for stability.
This paper aim at modelling of an ASMS and the application of the regularization method
in the problem of finding similar space images in the archives of the simulated ASMS by means of
discrete orthogonal transformations, in particular, wavelet transformations of Haar. We use the
Tikhonov regularization method, elements of mathematical analysis, the theory of discrete orthogonal
transformations, and methods for decoding cosmic images.
The result of the experiment, which confirms the Tikhonov regularization method for Haar
wavelet transforms, based on the example of processing archive satellite images that are located in a
data warehouse, is presented.
Introduction
In relation to the emergence of a variety of information technologies with
their numerous capabilities, wide-range monitoring has been provided for a variety
of phenomena and, accordingly, objects [1–9]. It is, first of all, the satellite systems
of Earth observation and large flows of information entering the scientific
laboratories. This led to the creation of new approaches and methods of organizing
71
the work with information on remote sensing of the Earth (RSE), as well as the
technologies for remote monitoring systems construction.
The major factors in the field of remote sensing that affect the development
of monitoring systems are:
In this paper, studies are carried out in the field of waste disposal facilities
(WDFs), which will be considered while creating a specific remote monitoring
system.
WDFs are a source of risk, associated with air and soil pollution, and pose
a real threat to the population and the environment [10–12].
"The big business gurus bequeathed to their descendants: the most
profitable investment is an investment in a person's natural needs. Among other
things, a person in the process of his life leaves behind mountains of debris. Only
the population of our country annually "produces" 35–40 million tons of solid
domestic waste"1.
It is almost impossible to control these objects, scattered on a giant
territory, by means of traditional land-based methods, and the annual use of aerial
observations becomes more and more expensive and can be carried out only in
limited local areas. Under these conditions, there is no alternative to space
1
http://www.rbc.ru/economics/23/08/2012/5703fbc99a7947ac81a6b045
72
monitoring as the most expeditious and cheap method for detecting and controlling
the development of unauthorized places for the WDFs.
The current article initiates a project on the creation of an automated space
monitoring system for the presence of WDFs in the territory of the Russian
Federation designed to provide an operative survey of the study area, the detection
and mapping of real and potential centres of unauthorized landfills, the forecast of
their development, a preliminary assessment of the scale of disasters and possible
consequences. It should be emphasized that such systems belong to the class of
open systems. They cannot be built in a complete form, they are in constant
development.
The initial project envisages the creation of a starter complex (core) of the
system, including the space and ground segments. An effective monitoring system
should comprise a space segment: spacecraft on circular solar synchronous orbits
equipped with active and passive sounding instruments of different spatial
resolution in the widest possible range of the electromagnetic spectrum. First of all,
it is advisable to use such public space systems such as NOAA (resolution 1100 m)
and EOS AM TERRA (resolution 250 m) for regular survey of state territory. Of
the high-resolution systems, the meteorological data "Meteor-3M" (37 m
resolution) and Indian satellites IRS 1С/1D (PAN resolution 5.6 m) are the most
accessible. It is these spacecraft that form the backbone of the space segment of the
system.
Since the research area is often restricted by a strong cloud cover, which
prevents regular space surveys in the microwave range, it is necessary to use active
sounding data, in particular, the Canadian satellite RadarSat-1, and Sentinel-1
(Copernicus). This satellite is equipped with radar, which allows obtaining high-
quality images of the Earth's surface, regardless of the presence of cloud cover and
time of day with a resolution of 8 to 100 m. In combination with space imagery in
the visible and thermal ranges, this information is extremely useful for detecting
and monitoring the development of unauthorized waste disposal facilities [9]. The
basis of the ground segment infrastructure of the ASMs WDF is the remote sensing
data acquisition system (RSAS). The most reasonable solution is to create
duplicate universal receiving stations and a distributed network for processing
space images. Creation of integrated automated WDF monitoring systems that
unite informational modelling and control systems, software complexes and
technical means of data collection and transmission on the basis of local computer
networks into a single structure, and the development of perspective models and
algorithms for forecasting foci, unauthorized WDF is an actual problem.
The functioning of space monitoring of waste disposal facilities (WDFs)
can be represented using the following scheme (Fig. 1).
73
Space monitoring of waste
disposal facilities
1 2 3 4 5
2
http://www.airsoft-bit.ru/stati-po-ekologii/394-groro
3
http:// www.ecobiocentre.ru/naturalist/google-svalka.doc
4
https://cyberleninka.ru/article/n/metodika-avtomaticheskogo-detektirovaniya-komponent-obektov-
zahoroneniya-othodov-po-kosmicheskim-izobrazheniyam
74
In addition to the above described system, it is possible to consider the
creation of a system for supporting long-term data archives5 along with and its
information input. These systems are actively created in recent years to address
both fundamental and applied problems. The main task of such systems is to
collect and organize long-term data storage and to conveniently represent this data
(import this data) into the information systems of various research projects.
We will consider the use of ready-made software for storage, output and
input of monitoring data as an alternative to creating software for such systems.
The most important thing of the inner workings of such systems is the electronic
Earth’s map of the site on Earth being under research. The electronic map has a
primary interface, it identifies the objects that are of interest to us, in this case,
these are solid waste objects, authorized and unauthorized, and sampling points,
through which soil survey is carried out, etc. The operation of the system will look
like this: some kind of GIS-system with open source. Then the map and data are
loaded into it.
The drawback of this approach is that GIS systems are either not adapted to
Russian or this is done poorly. This disadvantage introduces some discomfort in
the process of work, so it is advisable for one to conduct their own development of
ASMS SWF. When creating such an information system, you must have the
following components at hand: database management systems, a web server, a
programming language development environment and, of course, a web browser.
The question of obtaining an electronic map can be solved in two ways.
The first one is to use Yandex Maps or Google Maps services. You can optionally
download the map fragment of interest in a raster format and convert it to a vector
format, so you can create an interactive map, for example, using the Inkscape
application. This method has certain specificity – in spite of a fairly simple
manufacturing method, such a map does not provide information on the
composition of the soil of the investigated section of the earth.
The second method is the use of specialized maps on paper. Complete
information is provided, including soil data, etc. One could scan such maps and
convert them to vector format. However, this approach seems rather cumbersome
due to the fact that paper maps are not always in good condition and special
computer programs for their improvement have to be applied.
Thus, summarizing the above reasoning, we may conclude that for storing
and processing of the earth monitoring data, it is possible to adapt applications to
the tasks, leading to a significant reduction in time.
5
http://www.ntsomz.ru/ks_dzz/nkpoi/catalog_service
75
The next block on the information stream is the block for image
processing, i.e. space images6, 7. With the widespread implementation of digital
communication systems, the urgency of recovery problems solving, reduction of
additive noise and reconstruction of multidimensional signals, obtained with photo
and video cameras and transmitted through communication channels, is
increasing8. In practice, the images are often distorted by noise, which happens
during the formation and transmission of signals through the communication
channel. When digital images are obtained, the noise source can be a CCD
detector-spectrometer, as well as fluctuation processes in photo sensors9. The
results of image restoration are widely used in automatic signal processing
systems, in digital photo and video recording systems and machine vision. The use
of multimedia and television digital systems increases the urgency of solving the
image reconstruction problems in the reconstruction of static and dynamic two-
dimensional signals. Reconstruction of images is an important area of application
of modern digital information processing systems in obtaining a reliable estimate
for visual and especially for automatic analysis. In most cases, when solving the
reconstruction task, it is required to estimate the missing pixel values of images
and video sequences, and also to select the most "similar" section from a large
number of already existing ones. The solution of this task involves retouching and
restoring of the missing fragments of images when removing scratches, defects,
unnecessary inscriptions, etc.
Currently, the technical implementation of digital systems is intensively
developing, their speed and energy efficiency are increasing. In this case, methods
and algorithms for signal and image processing are used.
6
http://window.edu.ru/resource/028/76028/files/PosobieERS.pdf
7
http://files.lib.sfu-kras.ru/ebibl/umkd/54/u_course.pdf
8
http://www.dissercat.com/content/korrektsiya-tsifrovykh-kosmicheskikh-izobrazhenii-na-osnove-
verifitsiruyushchego-modelirovan
9
http://www.dissercat.com/content/rekonstruktsiya-smazannykh-i-zashumlennykh-izobrazhenii-
metodami-regulyarizatsii-i-usecheniy
76
Basic concepts
An important step in constructing a mathematical model of the concept of
"similarity" for storing space images is the measurement of the appearance of the
image boundaries using the Haar J-step discrete wavelet transformation (DWT)
[13–15].
As it is well known, one of the problems in image processing is to find a
way to efficiently represent an image or SI in a compact form. In modern practice
of spectral analysis, recently acquired signals of a special kind, namely, wavelets
have become quite popular. Since we are studying the SIs, we are primarily
interested in two-dimensional discrete wavelet transforms. For being ascertained
[16], the two-dimensional wavelet transforms are based on one-dimensional
wavelet transforms that do not depend on the number of rows and columns of the
image. By virtue of this rule, we consider, basically, the horizontal and vertical
directions of the wavelets. Specifically, Haar's wavelets [17, 18].
1. These are piecewise constant functions defined on intervals having
different scales and, furthermore, they take two values {–1; +1}.
2. The Haar's maternal wavelet, which has a unit scale and zero offset, is a
function that takes the value +1 on the interval [0; 1/2) and –1 on the interval
[1/2; 1).
The Haar transformation (HT) is one of the simplest basic wavelet
transforms [19]. Let f f1 ,, f N – the one-dimensional discrete signal. The HT
divides the signal into two components, the average and the difference. Suppose
we have a sub-signal a1 a1 , a2 ,, a N / 2 consisting of mean values. It is defined
as follows:
(1)
(2)
(3) a an , n ; d d n , n .
The first of these signals represents a coarsened copy of the original, and
the second is informative or detailed for the original signal.
(4) .
77
A similar partition can be performed with respect to the vector a. Let's
consider an example with numbers. Let
0 1 2 3
I =
4 5 6 7
(5)
8 9 0 1
2 5
3 4
Let us apply HT to the given matrix, using the principle of division by
rows, thus we obtain the following matrix:
1 5 1 1
(6) 1 9 13 1 1
2 17 1 1 1
5 9 1 1
We apply HT to the given matrix, using the principle of column separation
of the obtained matrix. We obtain the following:
10 18 2 2
(7) 1 22 10 2 2
2 8 8 0 0
12 8 0 0
(8)
(9) ,
Here
A – area, including information about the global properties of the analysed SI;
H – horizontal area, including information on the horizontal components of the
analysed SI;
V – vertical area, including information on the vertical components of the analysed
SI;
D – diagonal component, including information on the diagonal components of the
analysed SI.
When decomposing the image [20–24], the one-dimensional fast Haar
transformation (FHT) is applied first to the rows, and then to the columns of pixel
78
values of the input of the displayed matrix. As a result, we have a two-dimensional
wavelet transform. As a result, the SI is divided into four equal parts LL, HL, LH,
HH:
LL HL
LH HH
LL 3 HL 3
HL 2
LH 3 HH 3
HL 1
LH 2 HH 2
LH 1 HH 1
79
The inverse two-dimensional wavelet transform restores the lowest
quadrant recursively. We take for the initial function of the space image, the image
of the landfill (Fig. 4).
Consider a seven-step fibreboard using db1 (Haar). According to the
structure, the database in which information on the waste disposal site (WDS) is
stored [9] is presented as:
1. Common (name, type, location, comments, notes, etc.);
2. Geographical parameters (coordinates in different geographical
projections, adjacent and including administrative-territorial units (AU),
etc.);
3. Geometric (area, perimeter, litter concentration, accuracy of estimation,
etc.) and other parameters.
Photographs and images, links to Internet resources, the degree of danger, the
time of occurrence of t1 and the disappearance of t2 (their detection in images),
notes are also included.
Unauthorized data
detection unit
Visualization of Archive
data
80
location on the Earth surface in AB format, where A – latitude, B – longitude in
HHHMMSS.SSSS format (HHH – degrees, MM – minutes, SS.SSSSS – seconds;
4 characters S set the maximum geographical accuracy of detection). The classifier
stores the current WDS, i.e. those that exist at the current time t (t').
Each sequential WDS entered in the data model is assigned the sequence
number N, which connects the database, the classifier and the WDS archive.
Historically, WDSs that have disappeared by the time t2 < t are transferred from the
classifier to the archive, with each new WDS assigned the next sequence number.
The database stores information about all WDSs (current and historical). So, the
principle of the SIs storing has been described above. The work in the database,
storing the SIs, is carried out according to the following principle. Each SI is
described in a text document containing administrative data (date, time of taking a
picture, etc.). Also included are terms and phrases for key search. The documents
found are linked directly to the snapshot, which can then be viewed. This is a
normal, standard way of constructing a DBMS. However, to store the SIs this
method does not provide a complete adequate description. Oral description of the
SI, which must be found in the database, may not meet the expectations of the user.
The archive, formed on this principle, does not allow the search for "likeness" of an
image.
NGNs are systems that allow mechanisms to search for such images. They
are relevant in connection with the active use of modern multimedia technologies.
Hence, the purpose of the research is the development of the NGN project – the
system for storing the SI, a description, and in particular, the WDS.
Thus, the database management system of the SI operates according to the
following algorithm (Fig. 5):
A space image is entered;
A text document contains administrative information about the next digital
image, i.e. date and time of image reception, then terms and phrases that
characterize its content are entered, certain keywords are also input [9];
Input the structure vector of the image.
Let h, v, d be the horizontal, vertical, and diagonal components of the wavelet-
Haar transformation of the original image or SI.
Definition 1:
The structural vector of the image f is:
(10) xf 2
1h
, 12v , 12d , a1 ,, J2 h , J2 v , J2 d , a J ,
81
12v
where 12h , 12v , 12d are the variances [13], a j j 1, , J is the
12h
measure of anisotropy. If greater than 1, the structure is oriented vertically; if less
than 1, the structure is oriented horizontally.
Definition 2:
Let two pictures f 1 , f 2 be given. Images f 1 , f 2 are similar, if for,
0 f1 , f 2 where
x
4J
(11) f1 , f 2 x f1
x f2
i
f1
xi f 2
2
i 1
g
Calculation of the structural vector x ;
Sorting images f of the database on the principle:
x x
g f1
x xg f2
x g x f3
. As a result, a
"response directory" is formed;
The recipient image directory f1 , f 2 , f 3 , is given to the user.
To improve the qualitative set of the "response catalogue" it is necessary to use
the procedure of "relevant feedback" in accordance with the following algorithm:
Step 1. Let the user identify the first images of the recipient catalogue as
relevant (suitable) or irrelevant (unsuitable). Denote by – most not suitable.
Denote by – a set of suitable images;
Step 2. A new search for similar images starts, for this purpose we
introduce a new structural vector
xg x f mis x fi
x new
xg
x f mis
x fi
,
f i Arel
x
4J
2
where x g i
g
;
i 1
82
Step 3. The response directory is generated, viewed visually. If the result
suits the user – Stop, otherwise go to Step 1.
Definition 3:
The problem of determining a solution z R u from the space F with
respect to the initial data u U is said to be stable on the spaces (F, U) if for
0 0 , from the inequality U u1 , u2 ( ) follows that
F z1 , z 2 , where z1 R u1 , z 2 R u 2 ; u1 , u 2 U ; z1 , z 2 F .
Definition 4:
The problem of determining a solution z from a space F with respect to
"initial data" u from a space U is called a metric space (F, U) correctly posed on a
pair if the following conditions are satisfied:
1. For any there exists a solution z from the space F;
2. The solution is uniquely determined;
3. The problem is stable on the spaces (F, U).
Tasks that do not meet the listed requirements are called incorrectly
delivered.
83
position. Elements are numbered. Next, wavelet transforms are applied to all
columns. The image decomposition is shown in Fig. 2.
In general form, the similarity algorithm for two SIs has the following
form: Let X(i,j) i , j 1, K , N be the original image or SI. The process of forming
a distorted image can be represented as the output of some linear system.
The mathematical model of the process of forming a noisy image has the
form:
Xδ(i, j)
X(i, j) H + F S∙ST F-1 (i, j)
84
Description of the algorithm
We will carry out the experiment in the following way. Let's give a
mathematical statement of the problem of removing non-informative pixels of the
SI with regularization [13]. Let:
X(i,j) i , j 1,, N be the initial image with data dimensions , considered
as the result of some random process with certain properties;
F – discrete orthogonal transformation (Walsh, Fourier, wavelet-Haar, etc.),
F-1 – the inverse transformation;
S – a matrix of dimension choice mN of rank m, 1 mN,
R – a regularizing matrix in the task of processing the original image X (i, j)
defined as follows:
(1, )
(2, )
(14) .
R
.
.
( N , )
X i , j F01 S 0T S 0 F0 X i, j
~
(15)
85
Step 3: an "extrapolation" is carried out using the matrix ST i.e. on the
receiving side, the resulting matrix is complemented to the dimension N
(for example, all components except those selected are assumed to be 0).
Step 4: the resulting matrix undergoes an inverse F-1 transformation;
Step 5: Multiplication by a matrix of regularizing factors R is carried out.
As a result of these steps, the original vector is restored with errors:
The problem consists in choosing and therefore (n,) so that for the
given k the following condition is satisfied: 2 <<1. The choice of and (n,)
for the corresponding orthogonal transformations depends on the input data and the
transformation structure. The form of the matrix R is defined for the continuous
case and is used in a discrete interpretation of the image compression problem with
regularization. Table 1 gives a specific form of the matrix R for various
transformations.
The table does not indicate the exact values of . The question arises of
determining the values of , for which the conclusions of the experiment are valid.
The specific value of the regularizing parameter is determined experimentally
and is given in Table 2. Here, 1 is the lower bound; 2 is the upper limit of the
change .
86
A chart is shown on Fig. 7 where the arrangement of compression errors
with regularization 2 and without regularization 1 for the average experimental
parameters is presented.
cos
cosR
Furie
Walsh
Haar
FurieR
HaarR
WalshR
2 4 6 8 10 k
Conclusion
The project to create an automated space monitoring system for the
presence of a WDS on the territory of the Russian Federation, designed to provide
an operative survey of the studied area, the detection and mapping of real and
potential foci of unauthorized dumps, the forecast of their development, a
preliminary assessment of the scale of disasters and possible consequences is of
great urgency. The current article explores the project model. The use of the data
warehouse in the task of space monitoring of the WDS enables not only improving
of the performance of environmental authorities, but also to enhance the
monitoring effectiveness of the WDS (space, aerological and terrestrial).
References
1. Aristov, M. A. Monitoring of solid waste landfills and detection of spontaneous garbage
disposal according to space survey data. GeoProfil.ru. № 2. http://internetgeo.ru/
uploads/ journals/geoprofile0209/monitoring_tbo.pdf
2. Pogorelov, A. V., S. V. Dulepa, D. A. Lipilin. Experience of space monitoring of
landfills in the territory of the Krasnodar Territory. Geomatika 2013, № 4, pp. 64–
71. http://geomatica.ru/pdf/2013_04/2013_4_64-71.pdf
87
3. Avetisyan, D., R. Nedkov. Determining the magnitude and direction of land cover
changes in the semi-natural areas of Haskovo Region, Southeast Bulgaria.
Geoscience and Remote Sensing Symposium (IGARSS), 2015 IEEE International,
IEEE, 2015, pp. 4637–40. DOI:10.1109/IGARSS.2015.7326862.
4. Stankova, N., R. Nedkov. Monitoring forest regrowth with different burn severity using
aerial and Landsat data. Geoscience and Remote Sensing Symposium (IGARSS),
2015 IEEE International, IEEE, 2015, pp. 2166–69. ISBN:978-1-4799-7929-5,
ISSN: 2153-7003, DOI:10.1109/IGARSS.2015.7326233.
5. Stankova, N., R. Nedkov, I. Ivanova, and D. Avetisyan. Integration of multispectral and
SAR data for monitoring forest ecosystems recovery after fire. Proc. SPIE 10444,
Fifth International Conference on Remote Sensing and Geoinformation of the
Environment (RSCy2017), 104441J, 10444, SPIE, 2017, ISBN:978-9963-697-24-
3, DOI:10.1117/12.2277313
6. Nedkov, R. Quantitative assessment of forest degradation after fire using ortogonalized
satellite images from Sentinel–2. Comptes rendus de l’Academie bulgare des
Sciences, 71, 1, publishing house of BAS „Prof. Marin Drinov“, 2018, ISSN:
2367-5535
7. Nedkov, R. Normalized differential greenness index for vegetation dynamics
assessmenT. Comptes rendus de l’Academie bulgare des Sciences, 70, 8, Prof.
Marin Drinov Publishing House of Bulgarian Academy of Scinces, 2017, pp.
1143–46, ISSN:1310–1331.
8. Panayotova, D., R. Nedkov, M. Dimitrova, I. Ivanova, and M. Zaharinova. Eco-
Monitoringal Investigation of Forest end Land-Used Area and Agro-Climate
Characteristics in the Land of the Municipality of Kardzhali, Using Aerospace and
GPS Data. Journal Biotechnology & Biotechnological Equipment Volume 23,
2009 - Issue sup1: XI Anniversary Scientific Conference, 23, sup1,
Taylor & Francis, 2009, ISSN: 1310-2818, pp. 200–03, DOI:10.1080/
13102818.2009.10818400
9. Avetisyan, D., R. Nedkov, E. Velizarova, and D. Borisova. SAR and optical data in land
degradation processes estimation: a case study from Southeast Bulgaria. Proc.
SPIE 10428, Earth Resources and Environmental Remote Sensing/GIS
Applications VIII, 10428, SPIE, 2017, pp. 104281D-1-104281D-9. ISSN: 0277-
786X, DOI:10.1117/12.2277945.
10. Aleksandrovna, L. N., M. L. Glinsky, A. A. Zubkov, V. V. Danilov, and
L. G. Chertkov. Monitoring of the state of the subsoil on the LRW burial site.
http://www.atomic-energy.ru/articles/2012/02/20/31086
11. Brovkina, O. V. Remote monitoring of anthropogenic disturbances in the taiga zone of
the Northwest of Russia. PhD dissertation, 2011, 125 p.
http://www.dissercat.com/content/distantsionnyi-monitoring-antropogennykh-
narushenii-taezhnoi-zony-severo-zapada-rossii
12. Yannikov, I. M., M. V. Sleptsova, and K. K. Krivoshapkin. Organization of
biomonitoring of chemically hazardous objects and landfills with the use of
identification polygons. Bulletin of the North-Eastern Federal University. M.K.
Ammosov, NEFU, Yakutsk, 2013. pp. 15–25. http://cyberleninka.ru/
article/n/organizatsiya-biomonitoringa-himicheski-opasnyh-obektov-i-poligonov-
tbo-s-ispolzovaniem-identifikatsionnyh-poligonov
88
13. Kazaryan, M. L., M. A. Schahramanian, and A.A. Richter. Space monitoring of the
Earth and Haar wavelet transform. Future Communication Technology and
Engineering, Chapter 61, Edited by Kennis Chan CRC Press 2015, pp. 291–94.
Print ISBN: 978-1-138-02777-0. eBook ISBN: 978-1-315-69045-2. DOI:10.1201 /
b18331-65
14. Kazaryan, M. L., M. A. Schahramanian, and A.A. Richter. Multidimensional Haar
wavelet-transient research on the stability in the context of a satellite monitoring of
the Earth problem//Serbian journal of electrical engineering (in press)
15. Kazaryan, M., M. Shakhramanyan, R. Nedkov, A. Richter, D. Borisova, N. Stankova,
I. Ivanova, and M. Zaharinova. Research of the generalized wavelet
transformations of Haar correctness in remote sensing of the Earth//Proc. of SPIE
Vol. 10427-74 (Conference 10427: Image and Signal Processing for Remote
Sensing Conferences: 11–14 September 2017, Warsaw, Poland), p. 99.
https://spie.org/ERS/conferencedetails/image-signal-processing-remote-
sensing#227857
16. Richter A. A., M. L. Kazaryan, M. A. Shakhramanian, D. Borisova, N. Stankova, I.
Ivanova. Information modeling of waste disposal sites. Ecological Engineering
and Environment Protection, 1, 2017, pp. 15–21, URL: http://ecoleng.org
/archive/2017/1/15-21.pdf
17. Chui, C. K. Wavelets: A Mathematical Tool for Signal Analysis. SIAM, Philadelphia,
1997.
18. Daubechies, I. Ten Lectures on Wavelets. SIAM, Philadelphia, 1992.
19. Gonsales, R. C., R. E. Woods, S. L. Eddins. Digital Image Processing using Matlab,
Pearson Prentice Hall, Upper Saddle River, NJ, 2004.
20. Korolyuk, V. S., N. I. Portenko, A. V. Skorokhod, A.F. Turbine, et al. A handbook on
probability theory and mathematical statistics. M.: Science. The main edition of
physics and mathematics, 1985, 640 p.
21. Tikhonov, A. N. and V. Ya. Arsenin. Methods for solving ill-posed problems. M.:
Science. Home edition of physical and mathematical literature, 1979, 288 p.
22. Mallat, S. A wavelet tour of signal processing. N. Y.: Acad. Press, 1999, 851 p.
23. Nedkov, R. Orthogonal transformation of segmented images from the satellite
Sentinel-2. Comptes rendus de l’Acad. bulg. des Sciences, 70, 5, Prof. Marin
Drinov Publishing House of Bulgarian Academy of Sciences, 2017, ISSN:1310–
1331, pp. 687–92.
24. Anuj, B. and A. Rashid. Image compression using modified fast Haar wavelet
transform. World Appl. Sci. J. 2009. 7, 5. pp. 647–53.
89
ИЗСЛЕДВАНЕ НА СТАБИЛНОСТТА НА АЛГОРИТЪМ ЗА ПОДОБИЕ,
ИЗПОЛЗВАН ВЪРХУ СИСТЕМА ЗА СЪХРАНЕНИЕ
НА СПЪТНИКОВИ СНИМКИ ЧРЕЗ ХААР УЕЙВЛЕТИ,
СПОРЕД ТИХОНОВ
Резюме
Дистанционните изследвания на Земята позволяват получаването на
информация със средна до висока пространствена разделителна способност
от космически летателни апарати, с хиперспектрални инструменти на борда.
В настоящата статия е разработен модел на автоматизирана система
за откриване на сметища чрез космически наблюдения (АСОС). Един от
компонентите на модела е модулът за съхранение на космически снимки,
който позволява не само да се подобри ефективността на институциите,
занимаващи се с опазване на околната среда, но и също така да се увеличи
производителността на мониторинга на сметища, базиран на космически
изследвания, въздушни наблюдения и наблюдения от Земята. Освен това се
изследва проблемът за търсене на подобия в архиви с космически снимки
чрез използване на Хаар уейвлет трансформацията за стабилност.
Тук се използва регуляризационният метод на Тихонов, елементи на
математическия анализ, теория на дискретните ортогонални трансформации
и методи за декодиране на космически снимки.
Научна новост в разработката е предложеният модел на автома-
тизирана система за космически мониторинг, предназначена за откриване на
сметища. Задачата за обработка на архивирани космически снимки, и по-
конкретно търсенето на подобия в архивите от космически снимки, е из-
следвана от гледна точка на некоректно дефинирана задача.
90