Розроблення алгоритмів розпізнавання дефектів у структурі кодової послідовності на поверхні модуляційних дисків
This study investigates algorithms for detecting and localizing defects in code sequence structures on modulation disk surfaces. It targets small anomalies in lithographically patterned elements that can cause readout errors or reduced measurement accuracy. A multi-level image-processing model combi...
Saved in:
| Date: | 2025 |
|---|---|
| Main Authors: | , , , , |
| Format: | Article |
| Language: | English |
| Published: |
The National Technical University of Ukraine "Igor Sikorsky Kyiv Polytechnic Institute"
2025
|
| Subjects: | |
| Online Access: | https://journal.iasa.kpi.ua/article/view/351403 |
| Tags: |
Add Tag
No Tags, Be the first to tag this record!
|
| Journal Title: | System research and information technologies |
| Download file: |  |
Institution
System research and information technologies| _version_ | 1867334453804662784 |
|---|---|
| author | Manko, Dmytro Beliak, Ievgen Kryuchyn, Andriy Ishchenko, Ruslan Zavarzina, Valentyna |
| author_facet | Manko, Dmytro Beliak, Ievgen Kryuchyn, Andriy Ishchenko, Ruslan Zavarzina, Valentyna |
| author_institution_txt_mv | [
{
"author": "Dmytro Manko",
"institution": "Institute for Information Recording of NAS of Ukraine, Kyiv"
},
{
"author": "Ievgen Beliak",
"institution": "Institute for Information Recording of NAS of Ukraine, Kyiv"
},
{
"author": "Andriy Kryuchyn",
"institution": "Institute for Information Recording of NAS of Ukraine, Kyiv"
},
{
"author": "Ruslan Ishchenko",
"institution": "National Transport University, Kyiv"
},
{
"author": "Valentyna Zavarzina",
"institution": "National Transport University, Kyiv"
}
] |
| author_sort | Manko, Dmytro |
| baseUrl_str | http://journal.iasa.kpi.ua/oai |
| collection | OJS |
| datestamp_date | 2026-02-02T20:49:24Z |
| description | This study investigates algorithms for detecting and localizing defects in code sequence structures on modulation disk surfaces. It targets small anomalies in lithographically patterned elements that can cause readout errors or reduced measurement accuracy. A multi-level image-processing model combines Gaussi-an smoothing, adaptive thresholding, morphological operations, and contour-based segmentation. Processing stages are formalized as mathematical operators for reproducible implementation. Defects are characterized using perimeter- and area-based metrics, and their area distribution is approximated by a normal law. A spatial model computes defect centroids, enabling comparative quality as-sessment of disk samples. The software provides an interface for tuning thresh-olds, visualizing contours and defect-area plots, and exporting results. Tests on real defective disks confirm the method’s reliable detection of local structural violations and its suitability for diagnostic systems. |
| doi_str_mv | 10.20535/SRIT.2308-8893.2025.4.01 |
| first_indexed | 2026-02-08T08:06:09Z |
| format | Article |
| fulltext |
D.Yu. Manko, Ie.V. Beliak, A.A. Kryuchyn, R.M. Ishchenko, V.V. Zavarzina, 2025
Системні дослідження та інформаційні технології, 2025, № 4 7
TIДC
ТЕОРЕТИЧНІ ТА ПРИКЛАДНІ ПРОБЛЕМИ
ІНФОРМАТИКИ
UDC 621.535
DOI: 10.20535/SRIT.2308-8893.2025.4.01
DEVELOPMENT OF ALGORITHMS FOR DETECTING DEFECTS
IN THE CODE SEQUENCE STRUCTURE
ON THE SURFACE OF MODULATION DISKS
D.Yu. MANKO, IE.V. BELIAK, A.A. KRYUCHYN,
R.M. ISHCHENKO, V.V. ZAVARZINA
Abstract. This study investigates algorithms for detecting and localizing defects in
code sequence structures on modulation disk surfaces. It targets small anomalies in
lithographically patterned elements that can cause readout errors or reduced measure-
ment accuracy. A multi-level image-processing model combines Gaussian smoothing,
adaptive thresholding, morphological operations, and contour-based segmentation.
Processing stages are formalized as mathematical operators for reproducible imple-
mentation. Defects are characterized using perimeter- and area-based metrics, and their
area distribution is approximated by a normal law. A spatial model computes defect
centroids, enabling comparative quality assessment of disk samples. The software pro-
vides an interface for tuning thresholds, visualizing contours and defect-area plots, and
exporting results. Tests on real defective disks confirm the method’s reliable detection
of local structural violations and its suitability for diagnostic systems.
Keywords: modulation disks, automated inspection, code sequence, microstructural
anomalies, image preprocessing, morphological analysis, contour segmentation.
INTRODUCTION
The integration of automated surface inspection methods into the technological
workflow of optical and micromechanical components, particularly modulation
disks, plays a crucial role in ensuring the accuracy and reliability of photoelectric
measurement systems [1–3]. Previous studies have reported that the formation of
high-precision coded structures on transparent substrates using photolithographic
techniques is often accompanied by the emergence of local defects. These defects
may result from technological inaccuracies, residual stresses, or surface contami-
nation [4–6]. In response to the growing demands placed on the metrological per-
formance of encoding systems, the development of effective technical diagnostic
procedures for the detection of defects within code sequences at submicron struc-
tural resolution has become increasingly relevant.
Traditional inspection methods based on visual assessment and manual sur-
face marking of modulation disks are significantly outperformed by modern ap-
proaches utilizing computer vision systems (Fig. 1). These advanced systems en-
able automated processing of digital images, integration with production lines,
D.Yu. Manko, Ie.V. Beliak, A.A. Kryuchyn, R.M. Ishchenko, V.V. Zavarzina
ISSN 1681–6048 System Research & Information Technologies, 2025, № 4 8
and real-time adaptation to new requirements through the implementation of neu-
ral network algorithms [7]. The development of corresponding algorithms for the
structured analysis of modulation disks is considered a priority direction in ad-
vancing the technology of high-precision optomechanical component fabrication
and verification.
An analysis of scientific studies devoted to the automation of defect detec-
tion in binary structures formed during the photolithographic deposition of code
sequences reveals the active development of two principal approaches: classical
algorithmic solutions [7–10] and machine learning-based methods [7; 11–14].
The first category focuses primarily on traditional image preprocessing tech-
niques, including filtering, adaptive thresholding, segmentation, and morphologi-
cal transformations [8–10]. These methods allow for both the restoration of the
digital image matrix and the basic detection of structural anomalies. However,
such approaches exhibit limited adaptability to changes in illumination, local dis-
tortions, and micro-scale defects, which are common in coded patterns produced
by photolithographic processes. The second category of research emphasizes the
use of neural network architectures, particularly convolutional neural networks
(CNNs), autoencoders, and transformer-based models [11–14], which offer supe-
rior classification accuracy and enhanced generalization in the presence of in-
complete input data and high noise levels. Nevertheless, the implementation of
these solutions imposes substantial computational demands, often requiring
graphics processing units (GPUs) or tensor accelerators, which complicates their
deployment in software systems operating in real-time environments [11–14].
Furthermore, training neural network models necessitates the preparation of large
datasets of annotated digital images with labeled defects, which may be infeasible
in production settings with a limited number of representative samples. These
considerations highlight the need for a comprehensive methodology that com-
bines the efficiency of classical image processing algorithms, the flexibility of
machine learning techniques, and the optimization of computational resources.
Such an approach should aim to strike a balance between defect detection accuracy,
processing speed, and adaptability to real-world industrial operating conditions.
Fig. 1. Evolution of automated inspection tools for code sequences on modulation disks
Development of algorithms for detecting defects in the code sequence structure …
Системні дослідження та інформаційні технології, 2025, № 4 9
The aim of this study is to develop a mathematically grounded approach for
detecting defects in the structure of code sequences on the surface of modulation
disks by integrating image preprocessing techniques, morphological analysis, and
statistical interpretation of the results. The primary focus is placed on constructing
a comprehensive methodology based on thresholding and contour analysis, em-
ploying adaptive filters, shape moments, and area distribution approximation of
the detected defects. Given the constraints of computational resources and the
need for integration with embedded control systems, the study does not explore
the broad application of resource-intensive neural network models. Instead, it pro-
poses an efficient software-based algorithmic solution that prioritizes detection accu-
racy, processing speed, and feasibility for deployment in industrial environments. The
proposed model is designed to ensure the identification of local structural anomalies
within code sequences, with the potential for future enhancements tailored to the
specific characteristics of high-precision optomechanical components.
PROBLEM STATEMENT: DEFECT DETECTION IN THE BINARY
STRUCTURE OF A CODE SEQUENCE
The present study addresses the task of automatic defect detection in a binary
structure formed on the surface of a modulation disk as a result of photolitho-
graphic reproduction of a code sequence. The corresponding structure is com-
posed of a periodic or quasi-periodic set of elements, which are read by optoelec-
tronic sensors with high spatial resolution [4–6]. The occurrence of defects in
such structures—such as geometric distortions, fragmented damage, local darken-
ing, or bright artifacts—can lead to positioning errors, signal readout failures, and
degradation of the specified level of metrological accuracy.
The defect detection task is formalized as a process of digital image analysis,
where the code sequence is represented as a binary mask corresponding to a two-
dimensional matrix }1;0{),( yxBM , which contains pixel values obtained after
thresholding the input data. The input dataset, in turn, is defined as a grayscale
image matrix ]255;0[),( yxGI . The objective of the software algorithm is to
localize and classify regions that potentially deviate from the expected geometry
of the binary structure. To achieve this, a sequence of filtering and morphological
operations is applied, resulting in a set of contours }{ nC , where each ];1[ Nn
denotes a distinct object in the input image matrix, indicating a possible defect in
the binary sequence structure. For each contour nC , the corresponding area nS
and perimeter nP are calculated based on the number of points forming the con-
tour. A contour nC is classified as defective if its geometric parameters fall out-
side the empirically or calibration-defined thresholds: minS , xSma , minP and xPma ,
which are set according to the objectives of the inspection system (see Fig. 2).
Thus, the problem of defect detection in a code sequence structure is reduced to
the construction of a computational procedure capable of reliably localizing
anomalous regions based on geometric features of contours formed through mor-
phological image processing
The selected approach avoids the use of complex machine learning models
by implementing a software algorithm with controllable parameters, which can be
D.Yu. Manko, Ie.V. Beliak, A.A. Kryuchyn, R.M. Ishchenko, V.V. Zavarzina
ISSN 1681–6048 System Research & Information Technologies, 2025, № 4 10
adapted to the limited computational resources of the hardware platform within
the automated inspection system.
MATHEMATICAL MODEL FOR DEFECT RECOGNITION IN THE CODE
SEQUENCE STRUCTURE ON THE SURFACE OF MODULATION DISKS
The formalization of the defect recognition procedure within the code sequence
structure on the surface of a modulation disk is based on the development of a
mathematical model comprising the stages of digital image preprocessing, mor-
phological filtering, geometric contour analysis, and statistical evaluation of the
parameters of the detected objects. The proposed model describes image trans-
formations as a sequence of operations applied to the input image matrix and the
resulting binary mask, thereby ensuring algorithmic modularity, reproducibility of
results, and adaptability to specific application requirements.
At the first stage, the digital image matrix is converted into grayscale format,
which reduces computational costs and enables processing based on the bright-
ness values of each element ]255;0[),( yxGI . To reduce the negative impact of
high-frequency noise and eliminate digital artifacts that may be mistakenly identi-
fied as defects, a Gaussian smoothing procedure is applied. The Gaussian smooth-
ing method is based on the convolution of the image matrix with a two-
dimensional kernel G , which is mathematically formalized as:
.
2
exp
2
1
),( де), ,)(*(),(
2
22
2
ji
jiGyxGGIyxGIG
Fig. 2. Algorithmic flowchart for processing the digital image matrix for defect detection
Development of algorithms for detecting defects in the code sequence structure …
Системні дослідження та інформаційні технології, 2025, № 4 11
The parameters of the two-dimensional Gaussian kernel were selected to
ensure a balance between background smoothing and the preservation of image
components. Following the smoothing stage, adaptive thresholding is applied to
convert the image into binary form while accounting for local variations in
illumination. Each pixel ),( yxGIG is mapped to a corresponding value in the
binary image ),( yxBM based on the average brightness ),( yxGI within a local
neighborhood of size yx , which was set to 1111 pixels in this study,
while the threshold offset parameter was determined empirically and adjusted
using an interactive control element:
.),(),(при 0),(
,),(),( при 1),(
yxyxGIyxBM
yxyxGIyxBM
GIG
GIG
In the software implementation, inverse thresholding was applied, meaning
that the binary mask is interpreted with reversed polarity and is formalized as
1),( yxBM when ),(),( yxyxGI GIG . As a result of the aforementioned
transformations, a binary mask }1;0{),( yxBM is obtained, in which potentially
defective regions are highlighted as connected components with high contrast
relative to the background. This stage is critically important for ensuring the clear
formation of contours in the subsequent steps of morphological analysis of the
image matrix.
After adaptive thresholding is applied, the binary mask matrix may contain
residual noise, small-size artifacts, and structural distortions in the components of
visual objects. To improve the quality of defect detection, classical morphological
operations are used, allowing for the restoration of object shapes within the binary
image matrix and the stabilization of the subsequent contour analysis stage. The
fundamental morphological operation in this context is the morphological closing
operation (MCO), which is implemented by sequentially performing dilation and
erosion procedures on the binary mask matrix. The application of the closing
operation to the binary mask ),( yxBM is mathematically formalized as:
. )(),( MKMKBMyxBM MCO
where MK is the structural element that defines the shape and size of the mor-
phological window (Morphological Kernel, MK). In the software implementation
used in this study, a 55 pixel kernel was applied, with all elements set to
1),( yxMK . The closing operation enables the suppression of digital artifacts
that cause internal holes, contour breaks, and distortions in the overall shape of
visual objects. This is followed by the application of the morphological opening
operation, which is performed in reverse sequence:
. )(),( MKMKBMyxBM MOO
The opening operation, in turn, is intended to remove small-size artifacts
from the image that do not correspond to actual visual objects, eliminate isolated
noise, and preserve the core geometry of larger objects. Thus, the sequential ap-
plication of closing and opening operations enables the formation of a refined bi-
nary mask in which local defects have clearly defined boundaries without internal
breaks or extraneous artifacts. This is critically important for the accurate extrac-
tion of contours in the subsequent stage. The structural element of the kernel MK
plays a key role in the quality of the restored image matrix. The selected rectangu-
lar kernel of 55 pixels ensures symmetric filtering of digital artifacts and thus
D.Yu. Manko, Ie.V. Beliak, A.A. Kryuchyn, R.M. Ishchenko, V.V. Zavarzina
ISSN 1681–6048 System Research & Information Technologies, 2025, № 4 12
enables proper processing of both horizontal and vertical components. If neces-
sary, the shape and size of the structural element can be adapted according to the
specific characteristics of the defects.
After the morphological processing of the image, a refined binary mask is
formed, reflecting potential regions with structural anomalies. The next step is the
contour detection procedure (CDP), which identifies closed sequences of pixels
that define the boundaries of connected components. Each contour is treated as a
separate object that may correspond to a local defect. For each detected contour
}{ nC , containing ];1[ Nn points, the perimeter nP and area nS are calculated,
serving as the fundamental geometric features:
,)()(
,)(
2
1
2
1
2
1
1
11
1
mmmm
M
m
n
mmmm
M
m
n
yyxxP
yxyxS
n
n
where nM is the number of points in contour nC . The contour nC is classified as
containing a defect based on the threshold value pairs };{ mamin xSS and
};{ mamin xPP , if at least one of the following conditions is satisfied:
,
,
ma
min
xn
n
SS
SS
.
,
ma
min
xn
n
PP
PP
Thus, a controllable feature set is formed for each detected defect in the fol-
lowing form:
},,,{ nnnnn YXSPD ,
where },{ nn YX are the coordinates of the centroid of the corresponding contour
nC . The resulting set }{ nD serves as an analytical basis for subsequent visual and
statistical analysis. After classifying contours as defective based on geometric
criteria, a set of the areas of the detected objects }{ nS is formed. To analyze the
statistical characteristics of the distribution, the mean area nS and the standard
deviation S are estimated as follows:
.)(
1
,
1
2
1
1
nn
N
n
S
n
N
n
n
SS
N
S
N
S
The corresponding parameters make it possible to quantitatively characterize
the variability of the geometric properties of the defects and to identify the
presence of anomalous objects whose areas significantly deviate from the mean
level. To visualize the statistical distribution, a histogram of defect areas is
constructed and supplemented by a normal distribution approximation. In this
case, the probability density is modeled by the function:
.
2
1 22
)(
s
nn SS
S
S ef
Development of algorithms for detecting defects in the code sequence structure …
Системні дослідження та інформаційні технології, 2025, № 4 13
The parameters nS and S are considered maximum likelihood estimates
(MLE) for the normal distribution. The proximity of the empirical distribution toa
normal profile serves as an indicator of the homogeneity of the detected defect
class. Deviations from the normal distribution may indicate the presence of for-
eign objects or structural inhomogeneities.
SOFTWARE-BASED DEFECT RECOGNITION IN THE CODE SEQUENCE
STRUCTURE ON THE SURFACE OF MODULATION DISKS
To validate the functionality and effectiveness of the proposed approach, a soft-
ware implementation of the defect identification algorithm for the code sequence
structure of a modulation disk was developed. The corresponding software mod-
ule integrates stages of preprocessing, morphological filtering, contour analysis,
defect classification, and statistical evaluation of defect parameters. The user in-
terface provides interactive tools for adjusting thresholding and classification pa-
rameters and enables visualization of processing results, including graphical rep-
resentation of detected defects, histogram construction of defect areas, and tabular
output of coordinates and numerical characteristics.
The defect identification algorithm was implemented as a modular Python
application with a graphical user interface. The system architecture is based on
the principles of separating image processing logic, parameter control, result
visualization, and data export to external formats. This structure ensures
flexibility, scalability, and ease of modification for individual stages of the
algorithm. The software algorithm consists of three key components:
1. The graphical data processing module is responsible for the step-by-step
transformation of the input image matrix, including Gaussian smoothing, adaptive
thresholding, morphological filtering, contour detection, and the calculation of
geometric and statistical parameters of the detected objects. The core element is
the “ImageProcessor” class, which implements the main logic for binary mask
analysis and the formation of the defect feature set.
2. The Graphical User Interface (GUI) is implemented using the “Tkinter”
library. This component enables image loading, interactive adjustment of the
adaptive thresholding value, visualization mode switching, display of analysis
results, and result saving. The interface is divided into functional panels: the con-
trol panel, the visualization area, and the text fields for statistics and coordinates.
3. The result-saving mechanism enables the export of detected defects in
graphical PNG and tabular CSV formats. The defect mask, annotated image with
highlighted objects, and centroid coordinates can be saved as separate files for
further use in technical inspection systems or external analysis.
To implement the aforementioned functions, the following external libraries
were used: “OpenCV” for image loading, preprocessing, morphological opera-
tions, contour detection, and calculation of geometric parameters; “NumPy” for
vectorized data processing and basic statistical computations; “Matplotlib” for
generating histograms and visualizing the area distribution of detected defects;
“Pandas” for constructing tabular structures and exporting results in CSV format;
and “Scipy.stats” for approximating the area distribution using a normal distribu-
tion curve. The architectural design is based on a clearly structured separation of
D.Yu. Manko, Ie.V. Beliak, A.A. Kryuchyn, R.M. Ishchenko, V.V. Zavarzina
ISSN 1681–6048 System Research & Information Technologies, 2025, № 4 14
component functions, enabling both local testing of individual modules and inte-
gration of the system into a broader software environment for technical inspection
and machine-based analysis.
The defect detection algorithm is implemented using the methods of the
“ImageProcessor” class and automated through interaction with the graphical user
interface elements. The system’s operational logic involves the sequential execu-
tion of the following stages:
loading the grayscale image matrix, which reduces computational com-
plexity by excluding color components;
Gaussian smoothing with a fixed kernel to reduce noise levels and prepare
the image for thresholding;
adaptive thresholding using a local mean, with an adjustable offset pa-
rameter controlled via a slider;
morphological filtering, including a closing operation to eliminate internal
breaks and an opening operation to remove noise;
contour analysis to determine the geometric characteristics of connected
components (perimeter and area) and evaluate their compliance with predefined
threshold criteria;
classification of contours as defects based on whether their area or pe-
rimeter exceeds or falls below the specified threshold values.
The interface allows the user to modify key processing parameters in real
time, such as the threshold offset for adaptive image binarization, the minimum
perimeter value for classifying an object as defective, and a visualization mode
toggle that enables or disables the overlay of circles on the centroids of detected
defects. These parameters make it possible to adapt the algorithm’s sensitivity to
various lighting conditions, image scales, and defect types.
To evaluate the accuracy of the core functionalities performed by the modu-
lar Python application, verification was carried out using real microphotographs
of modulation disk surfaces. The processing results demonstrate the system’s abil-
ity to effectively detect defects originating from the photolithographic process by
isolating anomalous regions based on geometric and statistical criteria. Figs. 3–5 pre-
sent the processing outcomes for microimages of code sequence samples “1”, “2”,
and “3”, respectively, showing the original grayscale microimage, the binary
mask with overlaid contours (green contours indicate objects without defect fea-
tures, while white circular markers denote objects classified as defective), as well
as the histogram of detected defect areas with an overlaid normal distribution
curve and corresponding statistical data:
total number of detected defects;
average defect area;
average defect perimeter;
range of defect areas;
range of defect perimeters.
The histograms constructed based on defect areas characterize the structure
of the sample and visualize its variability. To approximate the empirical
distribution, a normal distribution model was applied using the parameters of
mean defect area and standard deviation. The obtained parameters represent
maximum likelihood estimates and reflect a distribution skewed toward lower
values, which is typical for defects associated with microcracks, scratches, and
Development of algorithms for detecting defects in the code sequence structure …
Системні дослідження та інформаційні технології, 2025, № 4 15
contamination particles. It should be noted that statistical indicators provide
insight not only into the number but also the nature of the defects. For instance, a
high standard deviation indicates significant variability in defect sizes, which may
suggest inconsistency in the technological process. Additionally, the coordinates
of defect centroids can be used for targeted adjustment of the photolithography
system and for initiating subsequent stages of detailed inspection.
Fig. 3 presents the results of applying the algorithm to code sequence sample
“1”, demonstrating the system’s capability to effectively localize both isolated
anomalies and small-scale digital artifacts, thereby enabling a comprehensive as-
sessment of the processed surface condition.
Fig. 4 shows the analysis results for code sequence sample “2”, indicating a
higher total number of defects but with a lower maximum area and less pro-
nounced dominance of a single large defect. This suggests a different nature of
structural disturbance in the binary code sequence compared to sample “1”,
potentially associated with dust or contamination deposition processes or
exposure instability in certain regions.
a b
c
Fig. 3. Processing results for code sequence sample “1”: a — original grayscale microimage;
b — binary mask with overlaid contours; c — histogram of detected defect area distribution
D.Yu. Manko, Ie.V. Beliak, A.A. Kryuchyn, R.M. Ishchenko, V.V. Zavarzina
ISSN 1681–6048 System Research & Information Technologies, 2025, № 4 16
Fig. 5 presents the processing results for code sequence sample “3”, which
contains high-contrast geometric structures and noticeable foreign inclusions.
Sample “3” is characterized by greater area dispersion and the presence of pro-
nounced macro-scale defects. This is confirmed by both the numerical character-
istics and the shape of the histogram, where the normal distribution curve exhibits
strong asymmetry. Such results indicate localized disruptions during fabrication
or damage incurred during operation.
The presented results confirm the stability and consistency of the algorithm’s
performance under varying input conditions, such as defect geometry, image con-
trast, and noise variability. Thus, the proposed approach demonstrates high sensi-
tivity to local structural anomalies while maintaining robustness against back-
ground artifacts and digital noise. The analysis of defect area distribution
histograms shows that the system can adapt to changes in the nature of damage
and maintain the reliability of statistical evaluation even in cases of asymmetric or
anomalous distributions. As part of future improvements, it is planned to extend
the algorithm by integrating machine learning classifiers for automatic defect type
identification, incorporating spatial context in the analysis of centroid distribution,
Fig. 4. Processing results for code sequence sample “2”: a — original grayscale microimage;
b — binary mask with overlaid contours; c — histogram of detected defect area distribution
a b
c
Development of algorithms for detecting defects in the code sequence structure …
Системні дослідження та інформаційні технології, 2025, № 4 17
and optimizing processing procedures for implementation on computational mod-
ules of embedded machine analysis systems operating in real time.
CONCLUSIONS
The article presents a comprehensive methodology for the automatic detection of
defects in the binary structure of a code sequence on the surface of modulation
disks, combining image preprocessing methods, morphological analysis, and sta-
tistical evaluation of the geometric characteristics of objects. A mathematical
model is proposed that describes the stages of smoothing, adaptive thresholding,
filtering, and contour detection, followed by classification based on area and pe-
rimeter. The developed software module provides a complete processing cycle of
the input image: from conversion into a binary mask to the visualization of de-
tected defects and construction of histograms with normal distribution approxima-
tion. The modular system architecture and the presence of a user interface that
allows adjustment of key parameters enable the adaptation of the program to vari-
ations in image quality, scale, and the nature of defects. Experimental verification
on samples of binary code sequences of modulation disks demonstrated the algo-
c
a b
Fig. 5. Processing results for code sequence sample “3”: a — original grayscale microimage;
b — binary mask with overlaid contours; c — histogram of detected defect area distribution
D.Yu. Manko, Ie.V. Beliak, A.A. Kryuchyn, R.M. Ishchenko, V.V. Zavarzina
ISSN 1681–6048 System Research & Information Technologies, 2025, № 4 18
rithm’s ability to detect both microdefects and local structural anomalies of con-
siderable area. The stability of results under varying processing parameters con-
firms the algorithm’s adaptability and its suitability for implementation in techni-
cal diagnostic systems under constrained computational resources. Further
extension of the software functionality is possible through the use of machine
learning classifiers, application of spatial contextual analysis, and integration with
real-time hardware platforms to enable autonomous monitoring.
Conflicts of interest. There are no conflicts to declare.
Acknowledgments. The authors express their deep gratitude to the National
Research Foundation of Ukraine for financial support under the project
No. 2023.04/0004.
REFERENCES
1. Y. Huang, Y. Yang, J. Liang, Z. Miao, M. Zhao, Y. Zheng, “An optical glass plane
angle measuring system with photoelectric autocollimator,” Nanotechnology and Preci-
sion Engineering, 2(2), pp. 71–76, 2019. doi: https://doi.org/10.1016/j.npe.2019.06.001
2. L. Lei et al., “A study on length traceability and diffraction efficiency of chromium
gratings,” Photonics, 11(3), 233, 2024. doi: https://doi.org/10.3390/photonics11030233
3. T. Wavrunek, S. Ball, Z. Gotto, B. White, “An Adhesion-based Alternative to Sol-
vent Processing in Microfabrication,” Proceedings of The National Conference On
Undergraduate Research (NCUR) 2020 Montana State University, 2020. Available:
https://libjournals.unca.edu/ncur/wp-content/uploads/2021/01/3238-Trevor-
Wavrunek-FINAL.pdf
4. I.V. Kosyak, D.Yu. Manko, Ie.V. Belyak, A.A. Kryuchyn, “Methodology for trans-
forming code sequences in accordance with the modulation disk coordinate system,”
(in Ukrainian), Electronic Modeling, 46(5), pp. 35–49, 2024. doi: https://doi.org/
10.15407/emodel.46.05.035
5. V.V. Petrov, А.А. Kryuchyn, Ie.V. Beliak, D.Yu. Manko, I.V. Kosyak, O.G. Melnik,
“Advantages of Direct Laser Writing for Enhancing the Resolution of Diffractive
Optical Element Fabrication Processes,” Physics and Chemistry of Solid State, 5(3),
pp. 587–594, 2024. doi: https://doi.org/10.15330/pcss.25.3.587-594
6. А.А. Kryuchyn et al., “Prospects for the creation of the technology of maskless pho-
tolithography based on direct laser recording,” Semiconductor Physics, Quantum
Electronics & Optoelectronics, 28(1), pp. 93–101, 2025. doi: https://doi.org/
10.15407/spqeo28.01.093
7. E.R. Davies, Computer vision: Principles, algorithms, applications, learning. Aca-
demic Press, 2018. doi: https://doi.org/10.1016/C2015-0-05563-0
8. X. Chen, “Optimization of image processing methods based on wavelet transform
and adaptive thresholding,” Applied Mathematics and Nonlinear Sciences, 9(1),
2023. doi: https://doi.org/10.2478/amns.2023.2.00665
9. E. Turajlic, E. Buza, A. Akagic, “Honey Badger algorithm and chef-based optimiza-
tion algorithm for Multilevel Thresholding Image segmentation,” 2022 30th Tele-
communications Forum (TELFOR). doi: https://doi.org/10.1109/telfor56187.
2022.9983775
10. D. Sundararajan, “Morphological image processing,” Digital Image Processing,
pp. 217–256, 2017. doi: https://doi.org/10.1007/978-981-10-6113-4_8
11. Z. Lyu, C. Zhang, M. Han, “A nonsubsampled countourlet transform based CNN for
real image denoising,” Signal Processing: Image Communication, 82, 115727, 2020.
doi: https://doi.org/10.1016/j.image.2019.115727
12. Z. Huang, H. Lu, X. Yu, H. Xiao, “Multi-Scale Feature Guided Transformer for Image
inpainting,” IET Image Processing, 19(1), 2025. doi: https://doi.org/10.1049/ipr2.70105
Development of algorithms for detecting defects in the code sequence structure …
Системні дослідження та інформаційні технології, 2025, № 4 19
13. P. Li, J. Chen, C. Cai, “Reinforced Res-Unet transformer for underwater image en-
hancement,” Signal Processing: Image Communication, 127, 117154, 2024. doi:
https://doi.org/10.1016/j.image.2024.117154
14. H. Wang, X. Lu, Z. Wu, R. Li, J. Wang, “Infrared and visible image fusion based
on Autoencoder Network,” IET Image Processing, 19(1), 2025. doi:
https://doi.org/10.1049/ipr2.70086
Received 26.06.2025
INFORMATION ON THE ARTICLE
Dmytro Yu. Manko, ORCID: 0000-0003-1848-2952, Institute for Information Recording
of NAS of Ukraine, Ukraine, e-mail: dmitriy.manko@gmail.com
Ievgen V. Beliak, ORCID: 0000-0001-9045-0782, Institute for Information Recording of
NAS of Ukraine, Ukraine, e-mail: beliak1312@gmail.com
Andriy A. Kryuchyn, ORCID: 0000-0002-5063-4146, Institute for Information
Recording of NAS of Ukraine, Ukraine, e-mail: kryuchyn@gmail.com
Ruslan M. Ishchenko, ORCID: 0000-0003-0158-4020, National Transport University,
Ukraine, e-mail: rm_ischenko@ukr.net
Valentyna V. Zavarzina, ORCID: 0009-0005-1666-3620, National Transport University,
Ukraine, e-mail: valazavarzina48@gmail.com
РОЗРОБЛЕННЯ АЛГОРИТМІВ РОЗПІЗНАВАННЯ ДЕФЕКТІВ У СТРУКТУРІ
КОДОВОЇ ПОСЛІДОВНОСТІ НА ПОВЕРХНІ МОДУЛЯЦІЙНИХ ДИСКІВ /
Д.Ю. Манько, Є.В. Беляк, А.А. Крючин, Р.М. Іщенко, В.В. Заварзіна
Анотація. Дослідження присвячено алгоритмам виявлення та локалізації де-
фектів у структурах кодової послідовності на поверхнях модуляційних дисків.
Воно спрямоване на невеликі аномалії в літографічно структурованих елемен-
тах, які можуть спричинити помилки зчитування або зниження точності вимі-
рювання. Багаторівнева модель оброблення зображень поєднує гауссове згла-
джування, адаптивне порогове визначення, морфологічні операції та
сегментацію на основі контурів. Етапи оброблення формалізовано як матема-
тичні оператори для відтворюваної реалізації. Дефекти характеризуються за
допомогою метрик на основі периметра та площі, а їх розподіл за площею ап-
роксимується нормальним законом. Просторова модель обчислює центроїди
дефектів, що дає змогу виконувати порівняльне оцінювання якості зразків ди-
сків. Програмне забезпечення надає інтерфейс для налаштування порогів, візу-
алізації контурів та графіків площ дефектів, а також експорту результатів. Тес-
ти на реальних дефектних дисках підтверджують надійне виявлення локальних
структурних порушень та придатність методу для діагностичних систем.
Ключові слова: модуляційні диски, автоматизований контроль, кодова послі-
довність, порушення мікроструктури, попереднє оброблення зображень, мор-
фологічний аналіз, контурна сегментація.
|
| id | journaliasakpiua-article-351403 |
| institution | System research and information technologies |
| keywords_txt_mv | keywords |
| language | English |
| last_indexed | 2026-02-08T08:06:09Z |
| publishDate | 2025 |
| publisher | The National Technical University of Ukraine "Igor Sikorsky Kyiv Polytechnic Institute" |
| record_format | ojs |
| resource_txt_mv | journaliasakpiua/54/838c106266e665599e83c470661dd954.pdf |
| spelling | journaliasakpiua-article-3514032026-02-02T20:49:24Z Development of algorithms for detecting defects in the code sequence structure on the surface of modulation disks Розроблення алгоритмів розпізнавання дефектів у структурі кодової послідовності на поверхні модуляційних дисків Manko, Dmytro Beliak, Ievgen Kryuchyn, Andriy Ishchenko, Ruslan Zavarzina, Valentyna модуляційні диски автоматизований контроль кодова послідовність порушення мікроструктури попереднє оброблення зображень морфологічний аналіз контурна сегментація modulation disks automated inspection code sequence microstructural anomalies image preprocessing morphological analysis contour segmentation This study investigates algorithms for detecting and localizing defects in code sequence structures on modulation disk surfaces. It targets small anomalies in lithographically patterned elements that can cause readout errors or reduced measurement accuracy. A multi-level image-processing model combines Gaussi-an smoothing, adaptive thresholding, morphological operations, and contour-based segmentation. Processing stages are formalized as mathematical operators for reproducible implementation. Defects are characterized using perimeter- and area-based metrics, and their area distribution is approximated by a normal law. A spatial model computes defect centroids, enabling comparative quality as-sessment of disk samples. The software provides an interface for tuning thresh-olds, visualizing contours and defect-area plots, and exporting results. Tests on real defective disks confirm the method’s reliable detection of local structural violations and its suitability for diagnostic systems. Дослідження присвячено алгоритмам виявлення та локалізації дефектів у структурах кодової послідовності на поверхнях модуляційних дисків. Воно спрямоване на невеликі аномалії в літографічно структурованих елементах, які можуть спричинити помилки зчитування або зниження точності вимірювання. Багаторівнева модель оброблення зображень поєднує гауссове згладжування, адаптивне порогове визначення, морфологічні операції та сегментацію на основі контурів. Етапи оброблення формалізовано як математичні оператори для відтворюваної реалізації. Дефекти характеризуються за допомогою метрик на основі периметра та площі, а їх розподіл за площею апроксимується нормальним законом. Просторова модель обчислює центроїди дефектів, що дає змогу виконувати порівняльне оцінювання якості зразків дисків. Програмне забезпечення надає інтерфейс для налаштування порогів, візуалізації контурів та графіків площ дефектів, а також експорту результатів. Тести на реальних дефектних дисках підтверджують надійне виявлення локальних структурних порушень та придатність методу для діагностичних систем. The National Technical University of Ukraine "Igor Sikorsky Kyiv Polytechnic Institute" 2025-12-29 Article Article application/pdf https://journal.iasa.kpi.ua/article/view/351403 10.20535/SRIT.2308-8893.2025.4.01 System research and information technologies; No. 4 (2025); 7-19 Системные исследования и информационные технологии; № 4 (2025); 7-19 Системні дослідження та інформаційні технології; № 4 (2025); 7-19 2308-8893 1681-6048 en https://journal.iasa.kpi.ua/article/view/351403/338422 |
| spellingShingle | модуляційні диски автоматизований контроль кодова послідовність порушення мікроструктури попереднє оброблення зображень морфологічний аналіз контурна сегментація Manko, Dmytro Beliak, Ievgen Kryuchyn, Andriy Ishchenko, Ruslan Zavarzina, Valentyna Розроблення алгоритмів розпізнавання дефектів у структурі кодової послідовності на поверхні модуляційних дисків |
| title | Розроблення алгоритмів розпізнавання дефектів у структурі кодової послідовності на поверхні модуляційних дисків |
| title_alt | Development of algorithms for detecting defects in the code sequence structure on the surface of modulation disks |
| title_full | Розроблення алгоритмів розпізнавання дефектів у структурі кодової послідовності на поверхні модуляційних дисків |
| title_fullStr | Розроблення алгоритмів розпізнавання дефектів у структурі кодової послідовності на поверхні модуляційних дисків |
| title_full_unstemmed | Розроблення алгоритмів розпізнавання дефектів у структурі кодової послідовності на поверхні модуляційних дисків |
| title_short | Розроблення алгоритмів розпізнавання дефектів у структурі кодової послідовності на поверхні модуляційних дисків |
| title_sort | розроблення алгоритмів розпізнавання дефектів у структурі кодової послідовності на поверхні модуляційних дисків |
| topic | модуляційні диски автоматизований контроль кодова послідовність порушення мікроструктури попереднє оброблення зображень морфологічний аналіз контурна сегментація |
| topic_facet | модуляційні диски автоматизований контроль кодова послідовність порушення мікроструктури попереднє оброблення зображень морфологічний аналіз контурна сегментація modulation disks automated inspection code sequence microstructural anomalies image preprocessing morphological analysis contour segmentation |
| url | https://journal.iasa.kpi.ua/article/view/351403 |
| work_keys_str_mv | AT mankodmytro developmentofalgorithmsfordetectingdefectsinthecodesequencestructureonthesurfaceofmodulationdisks AT beliakievgen developmentofalgorithmsfordetectingdefectsinthecodesequencestructureonthesurfaceofmodulationdisks AT kryuchynandriy developmentofalgorithmsfordetectingdefectsinthecodesequencestructureonthesurfaceofmodulationdisks AT ishchenkoruslan developmentofalgorithmsfordetectingdefectsinthecodesequencestructureonthesurfaceofmodulationdisks AT zavarzinavalentyna developmentofalgorithmsfordetectingdefectsinthecodesequencestructureonthesurfaceofmodulationdisks AT mankodmytro rozroblennâalgoritmívrozpíznavannâdefektívustrukturíkodovoíposlídovnostínapoverhnímodulâcíjnihdiskív AT beliakievgen rozroblennâalgoritmívrozpíznavannâdefektívustrukturíkodovoíposlídovnostínapoverhnímodulâcíjnihdiskív AT kryuchynandriy rozroblennâalgoritmívrozpíznavannâdefektívustrukturíkodovoíposlídovnostínapoverhnímodulâcíjnihdiskív AT ishchenkoruslan rozroblennâalgoritmívrozpíznavannâdefektívustrukturíkodovoíposlídovnostínapoverhnímodulâcíjnihdiskív AT zavarzinavalentyna rozroblennâalgoritmívrozpíznavannâdefektívustrukturíkodovoíposlídovnostínapoverhnímodulâcíjnihdiskív |