Системний аналіз проблеми встановлення справжності й авторства творів живопису
Cultural values have long been the objects of crimes, among which the export from the state stands out. Falsification hides artworks from customs control and its detection requires a long examination using a variety of methods of analysis. This article discusses the task of verifying painting’s auth...
Gespeichert in:
| Datum: | 2022 |
|---|---|
| Hauptverfasser: | , , , |
| Format: | Artikel |
| Sprache: | Englisch |
| Veröffentlicht: |
The National Technical University of Ukraine "Igor Sikorsky Kyiv Polytechnic Institute"
2022
|
| Schlagworte: | |
| Online Zugang: | https://journal.iasa.kpi.ua/article/view/250498 |
| Tags: |
Tag hinzufügen
Keine Tags, Fügen Sie den ersten Tag hinzu!
|
| Назва журналу: | System research and information technologies |
| Завантажити файл: |  |
Institution
System research and information technologies| _version_ | 1866302784865828864 |
|---|---|
| author | Martynenko, Andrii Tevyashev, Andriy Kulishova, Nonna Moroz, Boris |
| author_facet | Martynenko, Andrii Tevyashev, Andriy Kulishova, Nonna Moroz, Boris |
| author_sort | Martynenko, Andrii |
| baseUrl_str | http://journal.iasa.kpi.ua/oai |
| collection | OJS |
| datestamp_date | 2022-06-21T10:27:50Z |
| description | Cultural values have long been the objects of crimes, among which the export from the state stands out. Falsification hides artworks from customs control and its detection requires a long examination using a variety of methods of analysis. This article discusses the task of verifying painting’s authenticity during customs inspection. A two-stage procedure is proposed, which includes a quick check based on the analysis of painting’s images and a longer museum expertize. To implement the image analysis, it is proposed to use an intelligent decision-making system, which is based on a classifier that implements the k-nearest neighbors algorithm. A set of features to describe painting’s properties is formed, metrics for calculating the similarity measure on objects in the course of classification is proposed. To train an algorithm, a dataset is proposed, which includes paintings by world and European artists, as well as Ukrainian painters from different centuries. |
| doi_str_mv | 10.20535/SRIT.2308-8893.2022.1.04 |
| first_indexed | 2025-07-17T10:27:39Z |
| format | Article |
| fulltext |
A.A. Martynenko, A.D. Tevyashev, N.E. Kulishova, B.I. Moroz, 2022
50 ISSN 1681–6048 System Research & Information Technologies, 2022, № 1
UDC 303.732.4 + 004.67 + 004.8
DOI: 10.20535/SRIT.2308-8893.2022.1.04
SYSTEM ANALYSIS OF THE PROBLEM OF ESTABLISHING
THE AUTHENTICITY AND AUTHORITY
OF PAINTING WORKS
A.A. MARTYNENKO, A.D. TEVYASHEV, N.E. KULISHOVA, B.I. MOROZ
Abstract. Cultural values have long been the objects of crimes, among which the
export from the state stands out. Falsification hides artworks from customs control
and its detection requires a long examination using a variety of methods of analysis.
This article discusses the task of verifying painting’s authenticity during customs in-
spection. A two-stage procedure is proposed, which includes a quick check based on
the analysis of painting’s images and a longer museum expertize. To implement the
image analysis, it is proposed to use an intelligent decision-making system, which is
based on a classifier that implements the k-nearest neighbors algorithm. A set of fea-
tures to describe painting’s properties is formed, metrics for calculating the similar-
ity measure on objects in the course of classification is proposed. To train an algo-
rithm, a dataset is proposed, which includes paintings by world and European artists,
as well as Ukrainian painters from different centuries.
Keywords: intelligent decision-making system, automatic classification, k-nearest
neighbors, customs examination, paintings.
INTRODUCTION
The rapid rates of art market growth and constantly increasing demand for works
of fine art have led to the fact that the problem of authenticating works of art has
become extremely urgent for all market participants: art museums and galleries,
auctions, collectors and individuals, and for states customs services.
Works of fine art have long ceased to be just an expression of the artist’s
ideas and intentions; they often function as payment means and objects for profit-
able investments. For this reason, the paintings of famous masters have become
associated with criminal activities — forgeries, embezzlements, illegal transporta-
tion across state borders. The canvases of famous artists are especially widely fal-
sified. Falsifiers not only inflict enormous material damage on states, paintings
owners, but also spiritually devalue the works of great masters of painting, which
poses a threat to the economic security of states [1]. Falsification of works of fine
art means the production of counterfeit painting objects and their sale to obtain
material benefits. Depending on forger qualifications, used techniques, technical
means and materials in painting, there are fakes of different complexity — from
simple (copying) to super forgeries, to establish which authenticity is extremely
difficult even for specialists.
Expertize procedures are used to establish the paintings authenticity, or at
least to determine the degree to which they are classified as “cultural value” or
“national wealth”. There are two types of expertise — customs and museum ex-
System analysis of the problem of establishing the authenticity …
Системні дослідження та інформаційні технології, 2022, № 1 51
pertise [2]. The purpose of the customs examination is to ensure economic secu-
rity in the country. Customs expertise is strictly structured and has a hierarchy of
goals, where dating is at the top, and rest of data obtained is additional and basic
for the conclusion whether this work belongs to the appropriate category.
The ultimate goal of museum expertise is to establish the authenticity and
authorship of a painting. Currently, four main methods of authenticating pictures
are used: forensic, attributive, technological, complex [3].
The forensic method includes: studies of author’s signature on the picture;
examinations of painting author fingerprints; research of handwritten notes, signa-
tures, imprints of seals (stamps) on reverse side of the picture; analysis of prove-
nance reliability (the work ownership history from creation moment to present).
Currently, the concept of provenance has expanded: it also includes a list of
checks or invoices proving the fact of purchasing an item for a certain amount,
expert assessments, history of participation in auctions, reproductions in books
and catalogs, participation in exhibitions, as well as any references in relevant
literature.
The attributional method consists in studying the art form details to find out
the specifics of master individual style.
The technological method is implemented using various technical means of
analysis: microscopic, X-ray spectral, macrophotography, as well as photography
in ultraviolet and reflected infrared rays, etc. In the technological method of re-
search, all elements of picture are analyzed: base, soil, paint layer, etc. From the
obtained data, it is established that at various stages of his career, what certain
primers, paints, varnishes, brushes the artist used. The results show that each artist
has his own manner of “painting”, his own special technique, and style. To in-
crease reliability of making a decision on the paintings authenticity, a complex
forensic, technological and art history expertise is used [4]. The practical use of
complex examinations for paintings authentication requires not only involvement
of highly qualified experts groups equipped with the necessary technical means,
but also significant financial and time costs. A systematic solution to this problem
is possible based on intelligent video analytics, machine learning methods and
computational intelligence.
Taking into account the effect of time factor, we can say that the customs
examination should be more efficient to ensure a quick decision on the possibility
of exporting art object outside the state. Museum expertise is not so strictly lim-
ited in time, since it is not performed at border crossing points. Obviously, for
customs examination, the choice of methods used is rather limited in terms of
speed: these are variants of technological method based on photographing works
of art in different lighting conditions (studio, infrared, ultraviolet, X-ray), macro
photography. The rest of technological methods, as well as forensic, art history
and attributive expertise cannot be promptly performed under conditions of cus-
toms control. Therefore, a two-stage procedure for establishing the authenticity and
authorship of paintings is proposed (Fig. 1).
In this work, the first stage of the examination will be considered — the cus-
toms one, which is proposed to be implemented using one of technological meth-
ods, namely, photographing works of painting with high resolution under studio
lighting conditions.
A.A. Martynenko, A.D. Tevyashev, N.E. Kulishova, B.I. Moroz
ISSN 1681–6048 System Research & Information Technologies, 2022, № 1 52
ANALYSIS OF AREAS OF RESEARCH AND STATEMENT OF THE PROBLEM
Currently, research in this area is carried out in several directions. One direction is
more focused on the creation of new and improvement of existing devices that
allow the analysis of materials and substances that forms an object of art [1–6].
Another area is directly related to technologies of image digitization and
their analysis using statistics, signal processing, machine learning [7–10].
Deep neural networks, which have recently become popular, have also not
remained outside the attention of scientists who are developing means for com-
paring, identifying, and authenticating pictures. The most effective were convolu-
tional neural networks due to their ability to distinguish a large number of hetero-
geneous features in images [11–14]. A number of interesting solutions were found
during the use of generative networks [15–18].
This work is devoted to the problem of automatic identification of cultural
values, in particular, paintings using the intelligent decision support system
(IDSS) [19]. A system oriented to work in real time must have high speed and
high accuracy in solving the classification problem. For this, an approach based
on the weighted k-nearest neighbour’s algorithm is proposed, since for deep learn-
ing networks it is necessary to re-train if the value database is replenished by at
least one object.
METHODOLOGY
Formally, the classification problem can be presented as follows. There is a set of
n objects
nliiiiI nl ,1},,..,,....,,{ 21 ,
each of them is characterized by a set of m features
mqffffF mq ,1},,...,,...,,{ 21 .
Features take values from a certain set
phccccC qqqqq f
p
f
h
fff
,...,1 },,...,,...,,{
][][][
2
][
1
][
,
Fig. 1. Scheme of a two-stage expertize procedure for establishing the authenticity and
authorship of paintings
System analysis of the problem of establishing the authenticity …
Системні дослідження та інформаційні технології, 2022, № 1 53
where p is the number of possible discrete values of each feature.
One feature Tf is the target, its values for an objects set I make up a vector
Т
f СC T ][ . The classifier G learns by examples to establish relationships of
the form
TCIFG ))(( ,
calculating the approximated values of the target feature TĈ such that the differ-
ence between the specified and approximated values will be minimal:
min)ˆ,( TT CCd .
A trained classifier allows calculating target attribute values for new objects
,...},{ 21 nnnew iiI in this way:
TNewnew CIFG ))(( .
When identifying works of painting, the classification problem can be solved
for several target attributes [11]:
– determination of painting artistic style with target attribute TStyleC ;
– determination of picture genre with target attribute TGenreC ;
– defining the author with target attribute TArtistC ;
– determination of picture creation time with target attribute TTimeC .
Obviously, the dataset used to solve the problem must include the appropri-
ate attributes. If the artist’s name is an attribute required for such datasets, then
defining and marking up an art style requires the participation of highly qualified
art historians. The markup of the painting genre is an even more difficult problem,
so this attribute may not be present in all datasets, what should be taken into ac-
count when developing an intelligent decision-making system.
In this paper, it is proposed to solve the problem of classifying paintings by
the attribute of creation time TTimeC (Fig. 2).
As noted in [8], when marking up data, art critics often use information
about the author’s belonging to a particular artistic style — this increases the ac-
curacy of identification. Therefore, it is possible to single out global and local
characteristic features necessary to recognize the painting author and, accord-
ingly, the time of the painting (Fig. 3).
Time:
1880–1890
Fig. 2. Illustration of a painting classification system
Time:
1888–1890
1
Fig. 3. Scheme of paintings classification based on global and local features
A.A. Martynenko, A.D. Tevyashev, N.E. Kulishova, B.I. Moroz
ISSN 1681–6048 System Research & Information Technologies, 2022, № 1 54
During their creative activity, many artists have changed the artistic manner
of writing, moved from one style to another, so the use of signs that can character-
ize the artistic style will be useful. In the datasets that are currently used in the
development of automatic classification systems, the following styles are most
often considered: abstract expressionism, baroque, constructivism, cubism, im-
pressionism, neo-classical, pop art, post-impressionism, realism, renaissance, ro-
manticism, surrealism, symbolism [8]. In this work, the artistic style will be con-
sidered as one of the auxiliary attributes.
FEATURE EXTRACTION
To describe the general properties of a painting, data on color and structural prop-
erties inherent in the entire image are most often used. To form sets of such data,
many different algorithms and descriptors are used: wavelet transforms, Radon,
Hough, Fourier, Chebyshev transforms and their combinations [7]; Gabor filters;
Local Binary Patterns (LBP); SIFT detectors [20]; textural features; first 4 mo-
ments; multidimensional histograms; edge statistics features, etc. [8]. In particu-
lar, it is by the first 4 moments and by edges, the Impressionists’s works can be
classified unambiguously. Contour markers convey information about brush
strokes style that is specific to each artist, which makes Impressionism, stand out
in comparison with other styles.
Surrealist paintings can be described more informatively with the help of
contour and object statistics, which reflects the presence of significant “empty”
areas in their works.
In [7] it is noted that the use of color data in descriptors compilation in-
creases the classification accuracy by 18,1%. However, researchers most often
work with color data in RGB representation, since this color space describes sig-
nals from image capture and displaying devices. In this work, it is also proposed
to use the CIELab color space [21], since it is focused on the unambiguous de-
scription of visual stimuli in accordance with human vision.
Thus, the following set of global picture descriptors is proposed:
1. Local Binary Patterns (LBP) [22] for describing texture properties. In this
work, an LBP implementation is used within a neighborhood of 20 pixels and
with a radius of 2 pixels.
2. Color modification of LBP to describe the color properties of the texture.
LBPs are calculated in R, G, B and CIEL, CIEa, CIEb color channels. The results
are combined using concatenation to form a multivariate histogram for each image.
3. The first 4 points — mean, standard deviation, skewness, and kurtosis,
calculated in the directions of 0, 45, 90, 135 degrees. Moments are calculated
along the “stripes” in the image in several specified directions. A 3-bin histogram
is plotted for each obtained data vector.
4. Tamura’s textural features are roughness, contrast, directionality, linear-
ity, roughness and regularity.
The texture roughness characterizes main details dimensions that form
the image. Its estimate is based on average values calculation within
neighborhood of pixels:
i j
kk
jib
yxA
22
),(
),( ,
System analysis of the problem of establishing the authenticity …
Системні дослідження та інформаційні технології, 2022, № 1 55
where ),( jib is brightness of pixel with ji, coordinates; k is neighborhood size;
the texture roughness is then
xxyxAyxAyxE kkk ),,(),(),( .
The texture contrast is estimated based on fourth moment 4 relative to
mathematical expectation and variance 2 within neighborhood:
25,0
4)(
),(
yxСk ,
where
4
4
4
— kurtosis.
The texture directivity is estimated based on quantized edge directions histo-
gram )(aHdir :
p wa
dirppeaksk
p
aHaarnyxD )()(1),( 2 ,
where peaksn is peaks number; pa — peak angular direction; r — coefficient
that depends on quantization of angles levels pa ;
y
x
ap
arctan calculated with
Pruitt contour detector.
Linear similarity ),( yxLk is evaluated as average coincidence of edge direc-
tions that match in pixels pairs separated by a distance along the edge direction in
each pixel.
Texture regularity is a generalized feature defined as
)(1),( sslinelikenelitydirectionacontrastcoarsenessk ryxR ,
where sslinelikenelitydirectionacontrastcoarseness ,,, are standard deviations for each
feature.
Roughness summarizes contrast and roughness of texture as follows:
),(),(),( yxCyxEyxRoughness kkk .
5. Radon transform features, calculated for angles of 0, 45, 90, 135 degrees
and then combined into 5-pocket histograms.
6. Haralik’s textural features — contrast, correlation, entropy, energy and
homogeneity. They are calculated based on the contingency matrix
I
jpipIpp
jiP
)]()(|),[(
),( 2121 ,
where 21, pp are pixels belonging to image.
Then contrast will be defined as
ji
Н jipjiyxС
,
2 ),()(),( ;
the correlation is as follows:
ji ji
ji
Н
jipji
yxСorr
,
),()()(
),( ,
A.A. Martynenko, A.D. Tevyashev, N.E. Kulishova, B.I. Moroz
ISSN 1681–6048 System Research & Information Technologies, 2022, № 1 56
entropy:
i j
Н jipjipyxEntropy ),(log),(),( 2 ,
energy:
i j
Н jipyxEnergy 2),(),( .
7. Palette redundancy [9]:
max
max
H
HH
M RGB
B
,
where maxH is maximum image entropy, which for 8-bit color coding is
2438 ; RGBH — entropy calculated for individual BGR ,, channels.
It is proposed to describe features of individual areas of image, individual
details using local descriptors, which include:
1) euler’s number, minimum, maximum, median, mathematical expectation,
variance for each object;
2) SIFT descriptor, built for gray level images, and images in RGB and
CIELab color spaces.
The result of local features selection are descriptors in form of multidimen-
sional vectors.
CLASSIFICATION OF PICTURES IMAGES
A wide variety of artistic techniques, styles, colors that artists use when creating
paintings, leads, when developing an intelligent decision-making system, to the
need to implement such a classifier that will be highly adaptable as new samples
become available, and will also allow objects to be compared using a large set
very dissimilar features. In [7], a comparison of such classification methods as
weighted k-nearest neighbors method and SVM was made. The comparison
showed that weighted k-nearest neighbors method provides an increase in classi-
fication accuracy by about 15–20% compared to SVM.
Weights determination
To calculate weights, it is proposed to use two basic approaches:
1. Assigning weights in accordance with information gain criterion, which is
calculated on class-based entropy basis for attribute values:
C
c
cciE ppfW
1
2 )(log)( ,
where cp is number of objects with a feature value if and that belonging
to class c .
2. Assignment of weights in accordance with the Fischer method:
C
c
cc
C
c
cc
F
p
p
W
1
2
1
2)(
,
System analysis of the problem of establishing the authenticity …
Системні дослідження та інформаційні технології, 2022, № 1 57
where 2 , cc is mathematical expectation and standard deviation of data points
belonging to class c for a specific attribute; is global mathematical expectation
for all data points for a particular attribute.
Metrics used
To build a classifier, can use a variety of metrics, for example, Euclidean dis-
tance, Hamming distance, Mahalanobis, Minkowski, or Chebyshev distances.
However, the use of each metric is associated with peculiarities of internal data
structure or algorithm properties. These factors can lead to a significant decrease
in classification accuracy. The Chebyshev metric will be more useful when com-
paring objects by the same attribute. To determine Mahalanobis distance, it is
necessary to calculate observations covariance matrix, which for considered prob-
lem becomes is a laborious task. In addition, due to significant patterns data het-
erogeneity, the Mahalanobis metric will reduce classification accuracy due to co-
variance matrix “blurring” over entire data volume.
Thus, in this problem of classifying pictures using weighted k-nearest
neighbors algorithm, it is proposed to use Euclidean distance
jijninjiji ffffffffd 22
11 )(...)(),( ,
weighted Euclidean distance
22
111 )(...)(),( jninnjiji ffffffd .
Hamming distance
n
l
jlilji ffffd
1
),(
and the Minkowski distance
pn
l
p
jlilji ffffd
/1
1
),(
,
where ji ff , are vectors of attribute values for compared objects ji, ; jlil ff , — values
of l -th attribute for objects ji, being matched.
Obviously, all attribute values must first be normalized so that the condition
mlnifil ,...,1 ;,...,1 ],1,0[
now for each object рi , which is characterized by a vector of features if , the de-
gree of similarity with a certain class Tc is calculated in accordance with equation
n
i ci
ci
ci
T
T
T
d
d
M
1 ,
,
,
)(min
1
)(min
1
.
Dataset for experimental research
The activities of many museums now also include the digitization of stored valu-
ables to provide wider user access. Therefore, number of art databases now totals
dozens, each with thousands of images [11, 23–28]. There are databases contain-
A.A. Martynenko, A.D. Tevyashev, N.E. Kulishova, B.I. Moroz
ISSN 1681–6048 System Research & Information Technologies, 2022, № 1 58
ing mainly classical works. Others, on the contrary, are focused on contemporary
art. Still others represent different eras and artistic styles.
In this work, it is proposed to use a small-sized dataset [29] with open ac-
cess. This set contains works by 50 artists who worked at different times — from
the 15th to the 20th century. Their works are ranked among a variety of artistic
styles: Impressionism, Post-Impressionism, Northern Renaissance, Baroque, Ro-
manticism, Symbolism, Realism, Surrealism, Byzantine art, etc.
When studying the possibility of automatic identification of art values in our
country, it is important to take into account the historical context. It so happened
that for a long time access to works of world-famous artists was closed to
Ukraine, here it is much more likely that you can find paintings by Russian and
Ukrainian masters who worked in the 18–20 century in the appropriate artistic
style. Therefore, the set [29] was supplemented with images of paintings by Rus-
sian artists of the 17–19th century in the styles of romanticism, classicism, real-
ism [30]. These images were obtained from the official portal of the Hermitage
Museum, and data on the artists — in Wikipedia. It is important to note that a
considerable number of paintings are not attributed by author, but they also have
art value and may be subject to identification by style.
In addition, dataset also includes paintings by Ukrainian artists of the 19th
and 20th centuries, obtained from the portal of the National Art Museum of
Ukraine [31].
CONCLUSIONS
The paper considers the problem of paintings automatic identification using an
intelligent decision support system (IDSS). The authors proposed a solution in the
form of a classifier based on a weighted k-nearest neighbor’s algorithm.
The paper proposes a set of local and global features can be used to attribute
objects to be classified. The set of features includes color, texture, statistical and
other characteristics.
To calculate weights for k-nearest neighbors algorithm implementation, it is
proposed to use the Fisher method, as well as information gain criterion. In the
algorithm for similarity measure calculating, authors proposes to use several met-
rics: Euclidean, weighted Euclidean metrics, Minkowski and Hamming metrics.
As a dataset for experimental research, it was proposed to use a set includes
works by famous world and European artists, as well as supplemented by paint-
ings by Russian and Ukrainian masters.
REFERENCES
1. А. Mironova, F. Robache, R. Deltombe, R. Guibert, L. Nys, and M. Bigerelle, “Digi-
tal Cultural Heritage Preservation in Art Painting: A Surface Roughness Approach to
the Brush Strokes”, Sensors, 20, 6269, 2020. doi:10.3390/s20216269.
2. K. Zreik and N. Bouhai, “Post-digital design: The Hyperheritage project”, in Pro-
ceedings of the SHoCK—Sharing of Computable Knowledge; A. Fioravanti, S. Cursi,
S. Elahmar, S. Gargaro (Eds.); eCAADe 35: Rome, Italy, 2017, pp. 49–54.
3. W.S. Elkhuizen et al., “Comparison of three 3D scanning techniques for paintings, as
applied to Vermeer’s ‘Girl with a Pearl Earring’”, Herit. Sci., 7, 89, 2019.
4. M.J.W. Van Hengstum et al., “Development of a high resolution topography and
color scanner to capture crack patterns of paintings”, Proceedings of the
EUROGRAPHICS Workshop on Graphics and Cultural Heritage, Vienna, Austria,
12–15 November 2018, pp. 11–20.
System analysis of the problem of establishing the authenticity …
Системні дослідження та інформаційні технології, 2022, № 1 59
5. C.-C. Carbon and V.M. Hesslinger, “On the nature of the background behind Mona
Lisa”, Leonardo, 48, pp. 183–184, 2015.
6. G.J.A.M. Eumelen, E. Bosco, A.S.J. Suiker, A. van Loon, and P.D. Iedema, “A
computational model for chemo-mechanical degradation of historical oil paintings
due to metal soap formation”, J. Mech. Phys. Solids, 132, 103683, 2019.
7. L. Shamir, T. Macura, N. Orlov, M.D. Eckley, and I.G. Goldberg, “Impressionism,
Expressionism, Surrealism: Automated Recognition of Painters and Schools of Art”,
ACM Transactions on Applied Perception, vol. 7 (2), pp. 1–18, 2010.
8. F.S. Khan, S. Beigpou, J. Van de Weijer, and M. Felsberg, “Painting-91: a large
scale database for computational painting categorization”, Machine Vision and Ap-
plications, 25(6), pp. 1385–1397, 2014. Available: https://doi.org/10.1007/s00138-
014-0621-6
9. D. Wang and Y. Chen, “Calculation and Application of Xin’an Painting School Art
Style Model”, ICAITA, 1651, 012033, 2020. doi: 10.1088/1742-6596/1651/1/012033.
10. L. Liu, J. Chen, P. Fieguth, G. Zhao, R. Chellappa, and M. Pietikäinen, “From BoW
to CNN: Two decades of texture representation for texture classification”, Int. J.
Comput. Vis., 127, pp. 74–109, 2019.
11. B. Saleh and A. Elgammal, “Large-scale Classification of Fine-Art Paintings: Learn-
ing The Right Metric on The Right Feature”, Digital Art History, no. 2, 2016. Avail-
able: https://doi.org/10.11588/dah.2016.2.23376
12. V. Balntas, K. Lenc, A. Vedaldi, and K. Mikolajczyk, “HPatches: A benchmark and
evaluation of handcrafted and learned local descriptors”, IEEE Conference on Com-
puter Vision and Pattern Recognition (CVPR), 2017.
13. X. Zang, F.X. Yu, S. Karaman, and Sh.-F. Chang, “Learning discriminative and
transformation covariant local feature detectors”, IEEE Conference on Computer Vi-
sion and Pattern Recognition (CVPR), 2017.
14. K. He, Y. Lu, and S. Sclaroff, “Local descriptors optimized for average precision”,
IEEE Conference on Computer Vision and Pattern Recognition (CVPR), 2018.
15. R. Srinivasan and K. Uchino, Quantifying Confounding Bias in Generative Art: A
Case Study. Available: arXiv: 2102.11957v1 [cs.CY] 23 Feb 2021.
16. A. Elgamma, B. Liu, M. Elhosein, and M. Mazzone, “Can: Creative adversarial net-
works, generating ”art” by learning about styles and deviating from style norms”, In-
ternational Conference on Computational Creativity (ICCC), 2017.
17. M. Sung, The ai renaissance portrait generator isn’t great at painting people of
color. 2019. Available: https://mashable.com/article/ai-portrait-generator-pocs/
18. W.R. Tan, Ch.S. Chan, and K. Tanaka, “ArtGAN: Artwork synthesis with condi-
tional categorical GANs”, 2017 IEEE International Conference on Image Process-
ing (ICIP), 2017, pp. 3760–3764. doi: 10.1109/ICIP.2017.8296985.
19. A. Martynenko, V. Moroz, and I. Nulina, “An intelligent decision support system for
cultural property identification”, Computer-IntegratedTechnologies: Education,
Science, Production, 39, pp. 78–82, 2020. Available: http://cit-journal.com.ua/ in-
dex.php/cit/article/view/126
20. D.G. Lowe, “Object recognition from local scale-invariant features”, Proceedings of
the International Conference on Computer Vision, 2, pp. 1150–1157, 1999. doi:
10.1109/ICCV.1999.790410
21. M.D. Fairchild, Color Appearance Models. John Wiley and Sons, 2005, 340 р.
22. Z. Guo, L. Zhang, and D. Zhang, “A completed modeling of local binary pattern op-
erator for texture classification”, IEEE Trans. on Image Processing, vol. 19, issue 6,
pp. 1657–1663, 2010.
23. WikiArt. Visual art encyclopedia. Available: https://www.wikiart.org/en/
24. Arkyves. Available: http://arkyves.org/
25. Art UK. Available: https://artuk.org/
26. Artsy. Available: https://www.artsy.net/
27. Behance. Available: https://www.behance.net/
28. Artnet. Available: https://www.artnet.com/
29. Kaggle. Available: https://www.kaggle.com/ikarus777/best-artworks-of-all-time/tasks
A.A. Martynenko, A.D. Tevyashev, N.E. Kulishova, B.I. Moroz
ISSN 1681–6048 System Research & Information Technologies, 2022, № 1 60
30. Hermiage. Available: http://collections.hermitage.ru/entity/COLLECTION
31. National Art Museum of Ukraine. Available: http://namu.kiev.ua/ua/gallery/whole-
gallery.html
Received 15.01.2022
INFORMATION ON THE ARTICLE
Andrii A. Martynenko, ORCID: 0000-0002-5033-4696, Dnipro University of Technology,
Ukraine, e-mail: martynenko.andrey.a@gmail.com
Andriy D. Tevyashev, ORCID: 0000-0002-2846-7089, Kharkiv National University of
Radio Electronics, e-mail: andrew.teviashev@nure.ua
Nonna Ye. Kulishova, ORCID: 0000-0001-7921-3110, Kharkiv National University of
Radio Electronics, e-mail: nonna.kulishova@nure.ua
Boris I. Moroz, ORCID: 0000-0002-5625-0864, Dnipro University of Technology,
Ukraine, e-mail: moroz.boris.1948@gmail.com
СИСТЕМНИЙ АНАЛІЗ ПРОБЛЕМИ ВСТАНОВЛЕННЯ СПРАВЖНОСТІ
Й АВТОРСТВА ТВОРІВ ЖИВОПИСУ / А.А. Мартиненко, А.Д. Тевяшев,
Н.Є. Кулішова, Б.І. Мороз
Анотація. Культурні цінності давно є об’єктами злочинів, зокрема вивезення
їх із держави. Фальсифікація приховує твори живопису від митного контролю;
її виявлення потребує тривалої експертизи з використанням різноманітних ме-
тодів аналізу. Розглянуто завдання встановлення справжності картин під час
митної перевірки. Запропоновано двоетапну процедуру, яка передбачає швид-
ку перевірку на основі аналізу фотографій творів живопису та більш тривалу
музейну експертизу. Для реалізації аналізу фотографій запропоновано викори-
стовувати інтелектуальну систему прийняття рішень, дія якої базується на кла-
сифікаторі, що реалізовує алгоритм k-найближчих сусідів. Сформовано набір
ознак опису властивостей творів живопису, запропоновано метрики для обчи-
слення міри подібності об’єктів під час класифікації. Для навчання алгоритму
пропонується набір даних, що включає картини світових, європейських худо-
жників та українських майстрів різних століть.
Ключові слова: інтелектуальна система прийняття рішень, автоматична кла-
сифікація, k-найближчих сусідів, митна експертиза, твори живопису.
СИСТЕМНЫЙ АНАЛИЗ ПРОБЛЕМЫ УСТАНОВЛЕНИЯ ПОДЛИННОСТИ
И АВТОРСТВА ПРОИЗВЕДЕНИЙ ЖИВОПИСИ / А.А. Мартыненко, А.Д. Те-
вяшев, Н.Е. Кулишова, Б.И. Мороз
Аннотация. Культурные ценности давно являются объектами преступлений, в
частности их вывоз из государства. Фальсификация укрывает произведения
живописи от таможенного контроля; для ее обнаружения необходима дли-
тельная экспертиза с использованием разнообразных методов анализа. Рас-
смотрена задача установления подлинности картин в ходе таможенной про-
верки. Предложена двухэтапная процедура, которая предполагает быструю
проверку на основе анализа фотографий произведений живописи и более дли-
тельную музейную экспертизу. Для реализации анализа фотографий предло-
жено использовать интеллектуальную систему принятия решений, действие
которой базируется на классификаторе, реализующем алгоритм k-ближайших
соседей. Сформирован набор признаков для описания свойств произведений
живописи, предложены метрики для вычисления меры сходства объектов в
ходе классификации. Для обучения алгоритма предлагается набор данных, ко-
торый включает картины мировых, европейских художников, а также украин-
ских мастеров разных столетий.
Ключевые слова: интеллектуальная система принятия решений, автоматиче-
ская классификация, k-ближайших соседей, таможенная экспертиза, произве-
дения живописи.
|
| id | journaliasakpiua-article-250498 |
| institution | System research and information technologies |
| keywords_txt_mv | keywords |
| language | English |
| last_indexed | 2025-07-17T10:27:39Z |
| publishDate | 2022 |
| publisher | The National Technical University of Ukraine "Igor Sikorsky Kyiv Polytechnic Institute" |
| record_format | ojs |
| resource_txt_mv | journaliasakpiua/28/073a026aeec4b0f9b43880c0df3c2728.pdf |
| spelling | journaliasakpiua-article-2504982022-06-21T10:27:50Z System analysis of the problem of establishing the authenticity and authority of painting works Системный анализ проблемы установления подлинности и авторства произведений живописи Системний аналіз проблеми встановлення справжності й авторства творів живопису Martynenko, Andrii Tevyashev, Andriy Kulishova, Nonna Moroz, Boris интеллектуальная система принятия решений автоматическая классификация k-ближайших соседей таможенная экспертиза произведения живописи інтелектуальна система прийняття рішень автоматична класифікація k-найближчих сусідів митна експертиза твори живопису intelligent decision-making system automatic classification k-nearest neighbors customs examination paintings Cultural values have long been the objects of crimes, among which the export from the state stands out. Falsification hides artworks from customs control and its detection requires a long examination using a variety of methods of analysis. This article discusses the task of verifying painting’s authenticity during customs inspection. A two-stage procedure is proposed, which includes a quick check based on the analysis of painting’s images and a longer museum expertize. To implement the image analysis, it is proposed to use an intelligent decision-making system, which is based on a classifier that implements the k-nearest neighbors algorithm. A set of features to describe painting’s properties is formed, metrics for calculating the similarity measure on objects in the course of classification is proposed. To train an algorithm, a dataset is proposed, which includes paintings by world and European artists, as well as Ukrainian painters from different centuries. Культурные ценности давно являются объектами преступлений, в частности их вывоз из государства. Фальсификация укрывает произведения живописи от таможенного контроля; для ее обнаружения необходима длительная экспертиза с использованием разнообразных методов анализа. Рассмотрена задача установления подлинности картин в ходе таможенной проверки. Предложена двухэтапная процедура, которая предполагает быструю проверку на основе анализа фотографий произведений живописи и более длительную музейную экспертизу. Для реализации анализа фотографий предложено использовать интеллектуальную систему принятия решений, действие которой базируется на классификаторе, реализующем алгоритм k-ближайших соседей. Сформирован набор признаков для описания свойств произведений живописи, предложены метрики для вычисления меры сходства объектов в ходе классификации. Для обучения алгоритма предлагается набор данных, который включает картины мировых, европейских художников, а также украинских мастеров разных столетий. Культурні цінності давно є об’єктами злочинів, зокрема вивезення їх із держави. Фальсифікація приховує твори живопису від митного контролю; її виявлення потребує тривалої експертизи з використанням різноманітних методів аналізу. Розглянуто завдання встановлення справжності картин під час митної перевірки. Запропоновано двоетапну процедуру, яка передбачає швидку перевірку на основі аналізу фотографій творів живопису та більш тривалу музейну експертизу. Для реалізації аналізу фотографій запропоновано використовувати інтелектуальну систему прийняття рішень, дія якої базується на класифікаторі, що реалізовує алгоритм k-найближчих сусідів. Сформовано набір ознак опису властивостей творів живопису, запропоновано метрики для обчислення міри подібності об’єктів під час класифікації. Для навчання алгоритму пропонується набір даних, що включає картини світових, європейських художників та українських майстрів різних століть. The National Technical University of Ukraine "Igor Sikorsky Kyiv Polytechnic Institute" 2022-04-25 Article Article application/pdf https://journal.iasa.kpi.ua/article/view/250498 10.20535/SRIT.2308-8893.2022.1.04 System research and information technologies; No. 1 (2022); 50-60 Системные исследования и информационные технологии; № 1 (2022); 50-60 Системні дослідження та інформаційні технології; № 1 (2022); 50-60 2308-8893 1681-6048 en https://journal.iasa.kpi.ua/article/view/250498/255766 |
| spellingShingle | інтелектуальна система прийняття рішень автоматична класифікація k-найближчих сусідів митна експертиза твори живопису Martynenko, Andrii Tevyashev, Andriy Kulishova, Nonna Moroz, Boris Системний аналіз проблеми встановлення справжності й авторства творів живопису |
| title | Системний аналіз проблеми встановлення справжності й авторства творів живопису |
| title_alt | System analysis of the problem of establishing the authenticity and authority of painting works Системный анализ проблемы установления подлинности и авторства произведений живописи |
| title_full | Системний аналіз проблеми встановлення справжності й авторства творів живопису |
| title_fullStr | Системний аналіз проблеми встановлення справжності й авторства творів живопису |
| title_full_unstemmed | Системний аналіз проблеми встановлення справжності й авторства творів живопису |
| title_short | Системний аналіз проблеми встановлення справжності й авторства творів живопису |
| title_sort | системний аналіз проблеми встановлення справжності й авторства творів живопису |
| topic | інтелектуальна система прийняття рішень автоматична класифікація k-найближчих сусідів митна експертиза твори живопису |
| topic_facet | интеллектуальная система принятия решений автоматическая классификация k-ближайших соседей таможенная экспертиза произведения живописи інтелектуальна система прийняття рішень автоматична класифікація k-найближчих сусідів митна експертиза твори живопису intelligent decision-making system automatic classification k-nearest neighbors customs examination paintings |
| url | https://journal.iasa.kpi.ua/article/view/250498 |
| work_keys_str_mv | AT martynenkoandrii systemanalysisoftheproblemofestablishingtheauthenticityandauthorityofpaintingworks AT tevyashevandriy systemanalysisoftheproblemofestablishingtheauthenticityandauthorityofpaintingworks AT kulishovanonna systemanalysisoftheproblemofestablishingtheauthenticityandauthorityofpaintingworks AT morozboris systemanalysisoftheproblemofestablishingtheauthenticityandauthorityofpaintingworks AT martynenkoandrii sistemnyjanalizproblemyustanovleniâpodlinnostiiavtorstvaproizvedenijživopisi AT tevyashevandriy sistemnyjanalizproblemyustanovleniâpodlinnostiiavtorstvaproizvedenijživopisi AT kulishovanonna sistemnyjanalizproblemyustanovleniâpodlinnostiiavtorstvaproizvedenijživopisi AT morozboris sistemnyjanalizproblemyustanovleniâpodlinnostiiavtorstvaproizvedenijživopisi AT martynenkoandrii sistemnijanalízproblemivstanovlennâspravžnostíjavtorstvatvorívživopisu AT tevyashevandriy sistemnijanalízproblemivstanovlennâspravžnostíjavtorstvatvorívživopisu AT kulishovanonna sistemnijanalízproblemivstanovlennâspravžnostíjavtorstvatvorívživopisu AT morozboris sistemnijanalízproblemivstanovlennâspravžnostíjavtorstvatvorívživopisu |