=Paper=
{{Paper
|id=Vol-2677/paper14
|storemode=property
|title=Applied digital platform for remote sensing data processing
|pdfUrl=https://ceur-ws.org/Vol-2677/paper14.pdf
|volume=Vol-2677
|authors=Yuriy V. Avramenko,Roman K. Fedorov
|dblpUrl=https://dblp.org/rec/conf/itams/AvramenkoF20
}}
==Applied digital platform for remote sensing data processing==
https://ceur-ws.org/Vol-2677/paper14.pdf
Applied digital platform for remote sensing data
processing
Yuriy V. Avramenko [0000-0002-3082-1155] and Roman K. Fedorov [0000-0002-2944-7522]
Matrosov Institute for System Dynamics and Control Theory
of Siberian Branch of Russian Academy of Sciences, Irkutsk 664033, Russia
avramenko@icc.ru
fedorov@icc.ru
Abstract. New methods and techniques developed small research teams for
processing remote sensing that are rarely available to other teams. For the ex-
change of new methods among users, it is necessary to develop an applied digi-
tal platform that will allow publishing new methods, applying and comparing
the results obtained to choose the best one in solving their problems. This arti-
cle presents an applied digital platform being developed at the IDSCT SB RAS,
which allows the user to quickly receive and process remote sensing data for the
region of interest. The goals and requirements of creating an applied digital
platform are considered, a general architecture, individual components and ex-
amples of its application are considered.
Keywords: Remote Sensing, WPS, Landsat, Sentinel, Digital platforms.
1 Introduction
Since the data of remote sensing of the Earth appeared in the public domain, many
researchers began to use them to solve problems of thematic processing. In particular,
satellite images are widely used [1-2] for updating topographic and navigation maps,
agricultural monitoring, tracking the dynamics and state of forest felling, observation
of ice conditions and ets. It is difficult for users to choose an appropriate method or
technique among the well-known ones, since the result of their application depends on
the studied area. Often, in order to achieve a research goal, it is required to sort out
various combinations of processing methods and choose the most suitable one. The
traditional approach to processing remote sensing data is to apply a desktop applica-
tion which takes a lot of time. Desktop applications offer a limited set of methods,
which are sometimes not enough to solve the current task. Existing remote sensing
data processing systems, such as Google Earth Engine [3], ArcGIS Online [4], allow
you to create your own methods with limited development tools, and are not focused
on sharing them. New methods or techniques developed by research teams are rarely
used due to the complexity of distribution among users.
Therefore, it is promising to create an applied digital platform (ADP), within which
researchers could apply existing methods and techniques for processing remote sens-
Copyright © 2020 for this paper by its authors. Use permitted under Creative Com-
mons License Attribution 4.0 International (CC BY 4.0).
ing data, distribute their own methods, and compare the results. The goal of creating
an applied digital platform is to simplify the full cycle of remote sensing data pro-
cessing (see Fig. 1), i.e. automation of receiving, processing and publishing processes.
The ADP should provide tools for creating new methods based on software systems
or existing services.
Fig. 1. Workflow for remote data.
2 Applied digital platform
At the Matrosov Institute for System Dynamics and Control Theory SB RAS an ap-
plied digital platform is being developed for processing remote sensing data. The
ADP architecture is shown in the figure (see Fig. 2). ADP has a web user interface
accessible via a browser. Data processing is based on WPS services, which show
good results in standardization, software support [5], which makes it possible to cre-
ate and integrate custom services for each step of the remote sensing data processing
cycle. Let's consider the functions of the main components of the ADP.
The catalog allows the user to search for remote sensing data for the region of in-
terest. While searching users can specify the time interval, limit the cloudiness value,
select the sensor, etc. For each sensor the catalog offers a corresponding set of ser-
vices, processing and publishing.
The data storage system is designed for loading remote sensing data, storing in-
termediate results, processing results.
The service scheduling and execution system performs service execution, moni-
tors the state of execution, balances the load between computing nodes, determines
the schedule and sequence of service calls. The system has the ability to scale service
execution, authorize users for services, interact with the storage system, etc.
Relational data editing services are designed to create training dataset for some
supervised classification methods. The user through the browser can manually mark
the remote sensing data and specify the classes of objects. The results of the work are
saved in the form of relational tables with spatial attributes in the PostgreSQL DBMS
with the PostGIS module for working with spatial data.
Fig. 2. Application digital platform architecture.
3 Remote sensing data catalog
The catalog contains data from Landsat 7, Landsat 8, Sentinel-2 sensors from the
United States Geological Survey (USGS). Remote sensing data is updating every 5 -
16 days, depending on the sensors. Remote sensing data required for users is loaded
via the Google Cloud API using the developed script. The catalog stores meta infor-
mation about remote sensing data and allows users to search for remote sensing data
on the region of interest. While searching users can specify the time interval, cloudi-
ness, sensors (see Fig. 3).
Fig. 3. Remote sensing data catalog.
For each image, users can apply a set of service compositions that define the sequence
of service execution. The composition should contain a service for publishing data,
which will allow showing the processing result on the map. If the user needs to get
data that is not in the catalog, the user can start the WPS service for downloading
images.
4 Remote sensing data processing services
ADP allows publishing remote sensing methods in the form of WPS services. the
standard WPS implementation provides the following advantages:
• allows expanding the set of methods, ranging from simple spectral indices to
complex combinations of methods developed on different software platforms
and distributed around the world;
• unifies the application of methods through standardization;
• simplifies technical support (the service can be located on any geographical-
ly remote server and is available to the author-developer);
• significantly speeds up method applying due to the absence of the need to in-
stall software;
• services can be scaled to multiple servers when implemented in a cloud
ADP.
Remote sensing data processing techniques can be implemented using service
compositions that define the sequence of calls and parameters. Let's consider the
remote sensing data processing services implemented in the ADP and their fea-
tures.
Spectral index calculation service: NDVI, NDWI, NDSI, NDGI, NDMI, NPCRI,
PSRI, NBR, CMR, FMR, IOR, NDBI, AVI, BSI, SI (see Fig. 4). They are used for
monitoring large areas.
Clustering service based on calls to GRASS GIS and GDAL functions. Unlike the
desktop version, it was possible to reduce the number of steps required to get the
result from 11 to 3 (see Fig. 5).
Image segmentation service based on neural networks of well-known architectures
(Unet, FPN, Linknet, PSPNet), which allow for binary and multi-class segmenta-
tion (see Fig. 6).
Service for creating a training dataset. Data preparation for the training dataset is
performed using manual image marking. The user through the browser creates po-
lygonal objects with the class label, and at the output he receives a directory from
two subdirectories (images, ground_trust), if required, the training dataset can be
packed into a CSV file.
More details about these and other services can be found in [7].
Fig. 4. On the left - NDVI, on the right – NDSI.
Fig. 5. Monitoring changes. On the left - 2018, on the right – 2020 years.
Fig. 6. On the left - original image, on the right - segmentation result.
5 Data display services
Displaying data allows you to better evaluate the initial data and results, such tools are
developing in ADP [6]. Data display services allow users to view remote sensing data
in a browser on a map. For this, the Map service has been developed. The Map pub-
lishes the one image to the map, with styles passed as a parameter. The display of
remote sensing data is supported by a combination of channels, which allows you to
quickly obtain information, and can also be used to solve more complex problems.
Various combinations of channels are used to search for some objects in a satellite
image, for example, vegetation, urban areas, open soil, sandy areas and others. For
example, the 5-4-3 channel combination is known in the scientific literature as "artifi-
cial colors". A commonly used combination mainly for studying the condition of the
vegetation cover. The 4-3-2 channel combination "natural colors" corresponds to the
RGB color model (see Fig. 7). Other combinations are listed as channel numbers: 7-5-
3, 5-6-2, 5-6-4, 7-6-4, 7-6-5. To make the combination the WPS_visualize service
has been developed.
Fig. 7. Combination of channels. On the left "natural colors", on the right "artificial colors"..
6 Planning and Executing Service Composition
Processing and publishing services can be launched separately in a special form that is
generated based on the service metadata. In the remote sensing data catalog, images
are associated with service compositions leading to a finished result, i.e. the service
compositions must have publication service. Service composition is described in
JSON format. Each service of the composition has its parameter values. As a parame-
ter value, the user can define a value obtained from another service call. While start-
ing a service composition one image or a directory as a parameter value is attached to
the services composition, and then it is executed (see Fig. 8). JSON description of the
service composition is sent to the geoportal. The geoportal processes parameters,
converts them into link URLs, sets additional information, for example, access pa-
rameters, the address of the REST service to return the result, etc. Then the geoportal
transmits JSON to the service planning and execution system. The results of the work
are transferred to the catalog through the geoportal. The catalog displays the pro-
cessing results on a map.
Fig. 8. Executing service composition.
7 Conclusion
The developed applied digital platform allows the user to quickly receive and process
data on the region of interest. The use of the WPS standard makes it easier to incorpo-
rate new methods implemented for different operating systems using various devel-
opment environments. The use of service compositions, including receiving, pro-
cessing and displaying data, greatly simplifies the use of methods for processing and
analyzing remote sensing data.
8 Acknowledgment
The work was carried out with the support of RAS (projects: AAAA-A17-
117032210079-1, AAAA-A19-119111990037-0), RFBR (projects:18-07-00758-а,
17-57-44006-Mong-a) and Ministry of Science and Higher Education of the RF, the
grant for implementation of large scientific projects on priority areas of scientific and
technological development (project no. 13.1902.21.0033). Results are achieved using
the Centre of collective usage «Integrated information network of Irkutsk scientific
educational complex».
References
1. Silva Junior C.A. da Simulating multispectral MSI bandsets (Sentinel-2) from hyperspec-
tral observations via spectroradiometer for identifying soybean cultivars / Silva Junior
C.A. da, Teodoro L.P.R., Teodoro P.E., Baio F.H.R., Pantaleão A. de A., Capristo-Silva
G.F., Facco C.U., Oliveira-Júnior J.F. de, Shiratsuchi L.S., Skripachev V., Lima M., Nanni
M.R. // Remote Sensing Applications: Society and Environment – 2020. – Vol. 19 – №
March – P.1–14.
2. Bar S. Landsat-8 and Sentinel-2 based Forest fire burn area mapping using machine learn-
ing algorithms on GEE cloud platform over Uttarakhand, Western Himalaya / Bar S., Pari-
da B.R., Pandey A.C. // Remote Sensing Applications: Society and Environment – 2020. –
Vol. 18 – № March.
3. Google Earth Engine Homepage https://earthengine.google.com/ last accessed 2020/09/07.
4. ArcGIS Online Homepage https://www.arcgis.com/index.html last accessed 2020/09/07.
5. Lee K. A performance evaluation of a geo-spatial image processing service based on open
source PaaS cloud computing using Cloud Foundry on OpenStack / Lee K., Kim K. // Re-
mote Sensing – 2018. – Vol. 10 – № 8.
6. Yakubailik O.E. Software and technological support for visualization of satellite data /
O.E. Yakubailik, A.A. Kadochnikov, A.V. Tokarev // International research journal -
2018. - № 11 - P. 82-85.
7. Fedorov R.K. Services for analysing series of satellite images for estimating the change
dynamics of objects / Fedorov R.K., Rugnikov G.M., Avramenko Y. V. // CEUR Work-
shop Proceedings – 2019. – Vol. 2534 – P.267–272.