=Paper=
{{Paper
|id=Vol-2744/paper43
|storemode=property
|title=Some Aspects of a Client-Server Architecture System for Processing Radar Images
|pdfUrl=https://ceur-ws.org/Vol-2744/paper43.pdf
|volume=Vol-2744
|authors=Elena Chernetsova,Tatiana Tatarnikova
}}
==Some Aspects of a Client-Server Architecture System for Processing Radar Images==
https://ceur-ws.org/Vol-2744/paper43.pdf
Some Aspects of a Client-Server Architecture System for
Processing Radar Images
Elena Chernetsova1[0000-0001-5805-3111] and Tatiana Tatarnikova1[0000-0002-6419-0072]
1Russian State Hydrometeorological University, 79, Voronezhskaya st., St. Petersburg, Russia
chernetsova@list.ru
tm-tatarn@yandex.ru
Abstract. This article discusses some aspects of a client-server architecture sys-
tem designed to process radar images.It is assumed that data obtained remotely
are processed to determine oil pollution on the water surface.The synthesis of
monochrome images and the infological model of the system are considered.
The developed application provides the ability to preview the image, forming a
graphic file from satellite data; implements functions that allow you to annotate
images, marking areas of interest and adding comments; implements, if neces-
sary, an algorithm for merging monochrome images; implements a keyword
support system that allows flexible categorization of all images; provides the
necessary level of information security through the separation of user rights and
authorization systems. The developed software product allows access to files
stored in the GIS database archive in real time simultaneously by a large num-
ber of users, i.e. represents its network (web) application. The software product
contains three levels: a user interface on a client browser, a web application,
and a database server.
Keywords: Radar Image, Geographic Information System, Data Base, Mono-
chrome Image, Monochrome Image Merging Technique, Algorithm for Merg-
ing Monochrome Images Using an Information Measure, Radar Image Appli-
cation.
1 Introduction
Currently, the main source of information about the state of an extended monitoring
object (for example, water surface, land surface) are radar images obtained from Earth
satellites equipped with radar stations with a synthesized aperture (SAR) [1]. At pre-
sent, SARs are increasingly used in various technologies of remote sensing of the
Earth (RS), and in some of them, such as the study of dynamic processes in the ocean,
SAR is recognized as the only possible tool for obtaining reliable information. This is
due to two main circumstances that distinguish SAR from remote sensing sensors
operating in the visible and infrared ranges of the electromagnetic spectrum [2],[3]:
Copyright © 2020 for this paper by its authors. Use permitted under Creative Commons
License Attribution 4.0 International (CC BY 4.0)
2 E.Chernetsova and T. Tatarnikova
• SARs are capable of receiving radar images (RRI) of the Earth's surface, regardless
of the state of the cloud cover and surface illumination;
• The radar image is dependent on some specific characteristics of the underlying
surface: surface dynamics, dielectric constant, microrelief.
Accumulated SAR images are stored in aGeographic Information System(GIS) data-
base. The developed software product allows access to files stored in the GIS data-
base archive in real time simultaneously by a large number of users, i.e. represents its
network (web) application [4-6]
The software product contains three levels: a user interface on a client browser, a
web application, and a database server. Based on this, the system requirements for the
application under development are as follows:the interface is implemented using
HTML with support for JavaScript and Dynamic HTML based on templates, since
HTML is a means of organizing an interactive dialogue between the end user and the
levels of the system. JavaScript combined with DHTML makes the web page dynam-
ic. To develop a universal web application, a programming language is used that sup-
ports almost all used DBMSs. The DBMS, in turn, supports the standard SQL lan-
guage, which allows you to perform all kinds of queries: fetch, add, update, delete.
A prerequisite for providing access to the archive of SAR images is the ability to
preview the source file, that is, to obtain the so-called quicklook image. This opera-
tion must be fully automated. On an existing SAR image, it is necessary to be able to
note anomalies, add comments, i.e. Create annotated images.
An integral part of any system providing access to the database is the search en-
gine. When designing this application, a search engine was provided for several pa-
rameters: date, keywords. Search by date is possible both for a certain interval and for
a selected day. Keyword search implies the presence of a certain number of processed
images, that is, those to which the specialist assigned some categories. A single image
may contain several keywords, such as “slicks” and “temperature front”. In addition,
it is possible to also introduce an estimate of the frequency of occurrence of certain
phenomena in the pictures, which can be ensured by a simple calculation of keywords
by their frequency of occurrence. It is possible to expand the application and include
other geographical areas in the monitoring of the sea surface. A mechanism is provid-
ed for recording all user actions so that it is possible to undo any change.
It is possible to use the application simultaneously with a large number of visitors.
Application performance should not fall as the number of images on the site grows.
The interface is friendly, contains hints, has a concise, non-distracting design, naviga-
tion is convenient, supports hierarchical structure. Address (navigation) lines have a
convenient, catchy look.
The administrator responsible for maintaining the site has the following options:
• manage the application through its web browser;
• user management: editing data, changing a password, deleting, adding or removing
rights;
• adding and removing keywords;
• editing, adding and deleting user comments;
• creating and deleting user groups.
Some Aspects of a Client-Server Architecture System for Processing Radar Images 3
The developed system is located on a web server, accessible from anywhere in the
world around the clock and should at the first stage (during the first year) provide
access to at least 20 people a day with an average number of views of 30 pages with
SAR images.
A regular requirement for backing up data is also a necessary requirement.
2 Method of Monochrome Images Fusion
Usually, the observed sensor is somewhat biased and rotated relative to the reference
sensor, therefore, to increase the accuracy of determining the parameters of the im-
aged object, it is necessary to find a function F that more efficiently (with a minimum
deviation) displays the readings of the observed sensor S2 (x1, ..., xn) on the readings
of the reference sensor S1 (x1, x2, ..., xn) (Fig. 1).
F(S2)=S1
S2 S1
Fig. 1. Image Combination Method
The solution to the data fusion problem in this case is to find the translation vector
(rotation and displacement), which correctly calibrates two images with the same
configurations.
There are several ways to solve the problem of combining images. Consider the
“tabu search” (TS) and “genetic algorithms” (GA) [7],[8] algorithms used to integrate
the data of two images. The TS algorithm consists in translating one image relative to
another and finding a correspondence function between them. The points in which the
search was performed during the execution of the algorithm are stored in the “tabu”
list and are not viewed again while they are in it. The fast minimum is ensured by
optimization and the use of parallel search.
When executing the genetic algorithm, possible responses to queries are stored as
rows. Many of these rows form the gene pool. The quality of possible responses can
be evaluated through the fitness function. The relative quality of the answers is pro-
vided by the rows, which are used to create a new generation of rows, where their
contents are used to create new generation generations in which the contents of the
rows provide high-quality answers, more likely to continue the generation of the next
generation. There are many strategies for determining the operation of obtaining a
new generation.
4 E.Chernetsova and T. Tatarnikova
Each of the above algorithms has its own peculiarities: if TS is characterized by a
rapid convergence, as a rule, to local minima, then for GA it is a global minimum due
to a larger number of calculations. A new algorithm was proposed in [7], which has
the advantages of each of them. The proposed algorithm for the efficient registration
of two-dimensional images and their subsequent merging consists of the following
main steps.
1. The formation of the size of the "working frame" - the area of the sensors,
whichwill be subjected to translation and rotation, to determine the direction of
convergence of the algorithm.
2. Formation of the initial direction of convergence of the two sensor readings in-
herent in genetic algorithms in order to bypass local minima.
3. Choosing the direction and convergence step by maximizing the matching func-
tion Fitn_f is an action inherent in the TS algorithm.
4. Averaging the readings of the two sensors.
At the first stage, all necessary variables are initialized, and the readings of two sen-
sors (cameras) are obtained. The translation step, rotation angle, initial position and
size of the “working frame” of the image are also initialized.
At the second stage, a set of initial directions is generated along which the “work-
ing frame” is shifted in the observed image and for which the correspondence func-
tions between the reference and the observed images are calculated. The number of
generated directions is set empirically, and depends on the size of the processed im-
age. At the third stage, the direction with the maximum value of the correspondence
function is selected, and entered in the Tabu list. Then a cycle is organized, the num-
ber of iterations of which is established empirically depending on the degree of corre-
spondence of the two images and the achieved accuracy of their registration. Six pos-
sible directions (up, down, left, right, right and left angles of rotation) are selected for
broadcasting the “working frame”. The correspondence function is calculated for
these selected directions and they are searched in the Tabu list. If one of the directions
is already in the Tabu list, then the correspondence function of this direction is as-
signed the maximum value. Otherwise, the direction with the minimum value of the
correspondence function is determined and entered in the Tabu list. The selected di-
rection is used for further search. After completing this cycle, two images are merged
by translation in the selected direction of the observed image relative to the reference
and subsequent averaging of the two images.
Since the question of choosing the appropriate correspondence function was left
open in [9], it is proposed to choose the Kullback connectivity criterion [10] as the
correspondence function, which is nothing more than the amount of information on
Shannon in the image y about the image x
The Kullback connectivity criterion depends on the mutual probability distribution
density (PDD) of the pixel intensity of two images:
w ( x, y )
( x, y ) = log 2 w ( x, y ) (dxdy), (1)
w( x) w( y )
Some Aspects of a Client-Server Architecture System for Processing Radar Images 5
wherew(x,y)–mutual PDD of two images ;
w(x), w(y)– PDD of images x and images y respectively;
–measure of the distance between two pixels.
In [11], the choice of this function of the correspondence of two signals is justified,
since mutual information is one of the most accurate, powerful, and stable measures
of the correspondence of two signals due to the fact that
• There are no restrictions on the nature of the relationship between the pixel intensi-
ties of two images;
• No assumptions are made about the objects depicted; to apply the informational
connectivity measure, the image is not pre-parameterized or any of its characteris-
tics;
• Among all varieties of informational connectivity measures, the Kullback measure
has an almost linear loss function to minimize Bayesian risk when distinguishing
between two hypotheses [12].
The listed advantages of the selected informational compliance measure allow you to
fully automate the process of merging images without performing any preliminary
processing.
3 Infological Model of a Data Analysis System
Based on the task, we can distinguish the following main objects of the system: imag-
es; Comments marked areas; keywords; monitoring areas; users user actions; group of
users; user rights. The key object of the system is the “Images” array. The data in the
array comes from the data source. Arrays: Keywords, Comments, Actions, and
Marked Areas are related objects in the Image array. In turn, the “Image” belongs to
one of the “Regions”. The "Users" array is a subset of the "Groups" array. The objects
in this array are capable of creating related objects. Creating links is limited to the
"Rights" array.
To achieve maximum efficiency of data organization, any information array must
meet the requirements of the third normal form (which automatically includes the
requirements of the first two) [13]:
• each array entry must contain a unique code;
• each field of each record should depend on a unique code;
• the fields of the same name in all records in the case of duplicate data must contain
a unique code that refers to a separate directory.
The application should contain the following information about the stored SAR imag-
es: the path to the file (full file name), the path to the quicklook image, description of
the image, region, date of recording, date of receipt in the archive. The list of ele-
ments does not meet the requirements of the third normal form, so it is necessary to
6 E.Chernetsova and T. Tatarnikova
decompose. The object "district" is allocated in a separate directory with a unique
identifier (code).
Objects: “Comments”, “Keywords” and “Marked areas” belong to a separate im-
age and should contain the following information: image code, user code of the com-
menter, content, date of addition, date modified by the moderator or administrator, as
well as an activity flag (show on page or not). Here, the content can be the comment
text, or a keyword, or the coordinates of the marked area. The list of elements com-
plies with the requirements of the third normal form, since images and users are al-
ready listed in the corresponding tables.
The User Actions object must contain information about the action (add, disable,
delete or edit), the user code, the object over which the action was performed, and the
execution date.
All actions must be placed in a separate directory "Types of actions" containing the
action code, its name and description. Also, all system objects are listed in the "Ob-
jects" directory containing the object code and its name. Thus, the list of elements is
reduced to the third normal form.
Information about users. The application should contain a sufficient amount of in-
formation about the user in order to fully provide all the functionality. The following
data set allows this to be provided: user code, username, password, email address,
date of user addition, activation flag (access to the application is allowed or closed).
Here you can note the object directly associated with the user - this is a user group.
A separate table is selected for user groups containing the following information:
record code, group name. The user group object meets the requirements of the third
normal form.
The user rights object associates an individual user with his group. The user rights
table contains the following information: record code, user code, group code, and
meets the requirements of the third normal form.
As with any software product, erroneous or conflict situations may periodically oc-
cur in the application. To process errors and ensure stable operation of the application,
it is necessary to create a locking mechanism and error handling. The result is
achieved by combining the situation warning in the system software code and the
corresponding information in the database.
For these purposes, a “Services" object is required containing the following fields:
operation code, operation name, comment, execution flag, return code, creation date,
date of the last change.
Output information. To achieve the development goal, the web application must
have the following output:
- PCA image in preview mode (quicklook file), shooting date, short description,
file name;
• comments on the image;
• areas marked on the image;
• keywords.
The list and description of the output messages are presented in Table 1.
Some Aspects of a Client-Server Architecture System for Processing Radar Images 7
Input information. The input data for the application are SAR images in the form
of N1 format files, which are daily delivered to the server from rolling archives.
Table 1. List and description of output messages
Message Identifier Presentation Frequency Date of issue, s Recipient
of issue of infor-
mation
SAR- Images Videogram On demand Up to 120 un- User
image Mashinogram processed, 2-3
for processed
Comments Notes Array Same 2-3 Same
Marked selections Videogram Same 2-3 Same
areas
Keywords Tags Array Same 2-3 Same
N1 files can be of various types, depending on the shooting mode of the ASAR in-
strument:
• Image Mode (IM) – standard mode;
• Alternating Polarisation Mode (AP) –alternative polarization mode;
• Wide Swath Mode (WS) – broadband mode;
• Global Monitoring Mode (GM) – global monitoring mode;
• Wave Mode (WV) – spectral mode.
In addition, separate products are formed for each of the five shooting modes listed.
4 Results
To test the algorithm for merging monochrome images, we used a radar image of the
North Atlantic near the Galicia bank in the area of the “Prestige” tanker accident in
November 2002, obtained from the ERS-2 satellite [14]. Fig. 2 shows that for clear
sea water a weak contrast gray background is characteristic. Dark spots against this
background are surface films of oil pollution, white spots are sea vessels. From the
tanker “Prestige” - the white point in the southern part of the image, a dark plume
stretches northward, which is divided into two arms - the north and east. Separate
dark spots of the fuel emulsion, formed as a result of leakage from the “Prestige”
tanks in the first days of the accident, are clearly visible.
8 E.Chernetsova and T. Tatarnikova
Fig. 2. Source image
Fig. 3 shows the same image, but rotated and offset relative to the original. Fig. 4
shows the result of combining two images. Figure 5 shows the calculated mutual den-
sity distribution of the probability of pixel intensity. Figure 6 shows the mutual densi-
ty distribution of the probability intensities of the pixels, calculated in accordance
with expression (1) after applying the processing algorithm. Figure 7 shows the image
obtained as a result of the algorithm for merging monochrome images.
Some Aspects of a Client-Server Architecture System for Processing Radar Images 9
Fig. 3. Image rotated and offset from the original
Fig. 4. The result of combining two images
10 E.Chernetsova and T. Tatarnikova
Fig. 5. Mutual PDD of Combined Images
Fig. 6. Mutual PDD of two images that have undergone the fusion
Some Aspects of a Client-Server Architecture System for Processing Radar Images 11
Fig. 7. The result of monochrome images fusion
The above results indicate that the implementation of the algorithm for merging mon-
ochrome images at the parameter level using the information measure of compliance
allows you to fully automate the process of merging images without performing any
preliminary processing.
5 Conclusion
The developed application provides a preview only for accurate and medium resolu-
tion images in alternative polarization mode (APP, APM), accurate images in stand-
ard mode (IMP) and wideband images in medium resolution (WSM). This limitation
is due to the fact that these types of images come daily from the rolling archive, and
support for a larger number of products can greatly complicate the development and is
not essential.
The developed application provides the ability to preview images, forming a graph-
ic file from satellite data; implements functions that allow you to annotate images,
marking areas of interest and adding comments; implements, if necessary, an algo-
rithm for merging monochrome images; implements a key word support system that
allows flexible categorization of all images; provides the necessary level of infor-
mation security through the separation of user rights and authorization systems.
12 E.Chernetsova and T. Tatarnikova
References
1. Armand, N., Zakharov, A., Zakharova, L.: Synthetic Aperture Space Radars in Remote
Sensing of the Earth. In: Modern Systems and Future Projects. Earth exploration from
space, vol. 2, pp. 3-13 (2010).
2. Rosen, P., Hensley, S. et al.: Synthetic Aperture Radar In teleometry.In: Proc. IEEE, vol.
88(3), pp.333–382 (2000).
3. Sagues, L., Lopez-Sanchez, J. et al.: Polarimetric radar in terferometry for improved mine
detection and surface clutter rejection. In: IEEE Trans. Geosci. Rem. Sens, vol. 39, pp.
1271–1278 (2001).
4. Bogatyrev, V., Bogatyrev, S. , Bogatyrev A. Model and Interaction Efficiency of Com-
puter Nodes Based on Transfer Reservation at Multipath Routing. In: 2019 Wave Electron-
ics and its Application in Information and Telecommunication Systems (WECONF),
8840647 (2019).
5. Chernetsova, E., Shishkin, A.: Data processing and construction of oil pollution maps
based on radar observations. In: Materials of the international workshop “Integrated
Coastal Zone Management for Sustainable Development” under TACIS CBC Project No.
2005 / 100-731 “Tools for making management decisions in the integrated coastal zone
management system” , pp. 34-38 , St. Petersburg (2007).
6. Bogatyrev V., Bogatyrev S., Derkach A.: Timeliness of the Reserved Maintenance by Du-
plicated Computers of Heterogeneous Delay-Critical Stream. In: CEUR Workshop Pro-
ceedings, vol. 2522, pp.26-36 (2019).
7. Javad,S.,Saeid,S.,Seyed N., Akhavan,T.: Optimizing a hybrid vendor-managed inventory
and transportation problem with fuzzy demand: An improved particle swarm optimization
algorithm. Information Sciences, 272, 126–144 (2014).
8. Dziubenko, I., Tatarnikova, T.: Algorithm for Solving Optimal Sensor Devices Placement
Problem in Areas with Natural Obstacles. In: 2018 Wave Electronics and its Application in
Information and Telecommunication Systems (WECONF), pp. 1-5, St. Petersburg (2018).
9. Koval, V..: Algorithm of competitive fusion of sensory data in multisensor systems. Sen-
sors and Systems, 7, 39-41 (2002).
10. Kosenko, G.: Information Criteria for Distinguishing Signals. Radio and Communications,
Moscow (1982).
11. Chernetsova, E.: Remote monitoring of oil pollution in the aquatic environment. Publish-
ing House of the Russian State Humanitarian University, St. Petersburg (2008).
12. Simakhin, V.: Conditional measures of dependence.Bulletin of Tyumen State University.
Physical and mathematical modeling. Oil, gas, energy .7, 119-122 (2011).
13. Connolly, T., Begg, K.: Databases. Design, implementation and maintenance. Theory and
Practice. 3rd ed. Williams, Moscow (2003).
14. Groult, S.: Prestige oil disaster trial kicks off. IOL-News, 15.10.2012