Geoscan experience in creating a high-precision 3D model of the Tula region

Geoscan experience of the creation of the high-precision 3D model of the Tula region

Soloschenko F.V. – Head of Field Operations Department, Grinko E.V. – Head of Department of Real Estate Inventory and Cadastre, Kurkov M.V. – Head of Science and Research Department, Suzdalcev N.R. – Copywriter of “Geoscan” group of companies.

In recent years, one of the most rapidly growing fields in geoinformatics has been the three-dimensional terrain modelling, including urban spaces modelling. The main quality criteria in reconstruction of three-dimensional urban model are the accuracy of the model (the accurate positioning of landscape elements in the coordinate system), the realistic visualization of texture of building facades and structures, and the display of urban infrastructure elements. As 3D-modelling technology is developing, the range of applications for geospatial information is also growing, expanding from knowledge management (in tourism sphere, for example) to cartography, engineering, monitoring, etc. Illustrative examples of such projects are the high-precision models of Singapore and Helsinki [1]. The first project of 3D city modelling in the Russian Federation was carried out in 2014 by a group of companies "Geoscan" on the territory of Tomsk city. Following an aerial survey made by unmanned aerial vehicles (UAV), 190 000 images of the survey area (> 320 km2) with a resolution of 3-5 cm/pix were obtained. Based on these data, a 3D model, an orthomosaic, a digital terrain model (DTM) and photo panoramas of the city were created [2].

This experience has demonstrated all the advantages of using unmanned aerial vehicles for obtaining digital spatial information, which impulsed the launching of the National Technology Initiative project "Creation of a geodetically accurate 3D model of a pilot region in the Russian Federation based on unmanned aerial survey data and GLONASS technologies". The Tula region was chosen as the pilot region.

Fig. 1 – Geoportal of the Tula region.

The Federal Programme "Development of a Unified State Registration System of Rights and Cadastral Registration of Real Estate Objects (2014-2020)" [3] stimulated the registration of the previously unregistered real estate objects and acted as a catalyst for the initiative. However, the traditional cadastral technologies do not allow to determine the boundaries of more than 30 mln land plots in short time for further registration. The most effective solution to this problem, based on the use of unmanned aerial systems, was proposed by the specialists of Geoscan. This technology makes it possible to create orthophotos of the territories with a mean squared error in locating control points of land plots boundaries, as well as contours of buildings and structures, of less than 10 cm, and for agricultural lands – of less than 20 cm, which corresponds to the requirements specified by the Ministry of Economic Development of Russia Directive № 90 of 01/03/2016 [4].

During the project implementation, a complex of geodetic, aerial photographic, photogrammetric and analytical works was carried out on the territory of the Tula region, and the following tasks were accomplished as a result:

high-precision 3D model of the region's territory, orthomosaics, digital terrain models (DTM) have been created;
high-precision detailed models of cities and particular objects of cultural heritage have been produced for solving administrative and municipal problems;
technical, cadastral errors and errors of the land registration have been identified;
unused or inappropriately utilized agricultural lands have been identified;
the basic layers of the spatial data infrastructure of the Tula region have been set up;
the Geoportal of the Tula region has been created (Fig. 1);
the accuracy of the local reference surveying network of the Tula region has been tested;
an interface of the workplace for cadastral engineers performing cadastral works has been created.

Refinement of the technologies of the field works on the territory of the test regions

The territory of Zaoksky district of the Tula region became a test site, where, since June 2016, different methods of performing field works for the project have been tested. As a result, optimal methods for creation of the basis of a compilation survey for the aerial photographing were developed, which implies the positioning of reference base stations (RBS) during the period of operation in the area, relative to which the coordinates of control points (CPs) and base flight stations (BFS) are measured. The precise coordinates of the RBSs were determined by 4 points of the State Geodetic Network (Gosudarstvennaya Geodezicheskaya Set’ - GGS) and 5 points of the State Leveling Network [5]. Due to the poor state of the GGS points, their number was increased to 6, in order to obtain reliable results. For the accuracy control of the created orthomosaics, a large number of control points - 910 – were placed on the ground throughout the entire region. Later, it became clear that such an amount of CPs is redundant.

Based on the results of the works and refinements in the test areas, which also included Alexinsky and Yasnogorskiy regions and Novogurovsky urban district, studies were carried out which showed that the accuracy of coordinates of geodetic network points of the region was not homogenous, whereas it should have been consistent for the entire area within the space of the coordinate system MSK-71.1. An attempt to equalize the geodetic network of three regions in a single system of coordinates was unsuccessful (the deviations of the calculated values of the coordinates reached 30 cm). Therefore, it was decided to continue carrying out the work, basing on the points of the geodetic network of a particular area (preserving the relation with the points of neighboring areas).

In total, 4 reference base stations and 58 base flight stations have been installed in the test areas, coordinates of 1931 control points have been determined.

94 survey points of the local Reference Boundary Network (Opornaya Mezhevaya Set’ - OMS) were surveyed, and it was discovered that not all settlements were provided with the OMS points. As it can be observed from the graph in Fig. 2, the discrepancies between the measured and cataloged coordinates of the OMS points of the Zaoksky district reach more than 0.5 m, and the joint equation of the network is impossible. In addition, the survey showed that the overall state of the points of the network is deplorable, so it was decided not to use them.

Fig. 2 – the graph of the OMS points inspection of the test district. On the X axis – OMS points numbers of the test district. On the Y axis – mean square error, m, of the measured and cataloged values of the OMS points of the test district.

One of the objectives of the project was the creation of orthomosaics within the frame of the state system of coordinates of 2011 year (GSK-2011). The joint equation of the GGS points within the test areas showed promising results - the deviations of the calculated values of the coordinates did not exceed 3 cm. However, further, it was decided against performing the compilation survey for aerial photographing until changes are made in the current legislation, allowing the registration of objects and carrying out changes in cadastral configuration of the regions within the frame of GSK-2011.

An aerial survey of the Zaoksky district (area - 1310 km2) took 28 flight days. For these purposes, initially, 3 UAVs were used – two Geoscan 101 and one Geoscan 201 (the main difference between them is the flight duration - 1 and 3 hours, respectively), later - three Geoscan 101 and two Geoscan 201. The whole work package was carried out by three teams: two teams (composed of 2 people) were creating the basis for a compilation survey and one team (composed of 4 people) was conducting aerial surveying. The works were carried out in other test areas in a similar way. In total, it took 74 flight days to survey the area of 4037 km2.

Field works organization on the territory of the Tula region in 2017.

During the winter period (2016-2017), the results of the aerial survey in the test areas were analyzed. In order to improve the efficiency of field works, a number of changes were introduced:

the flight planning program was modernized, which made it possible to use one flight station to operate the flight of two UAVs simultaneously;
the number of Geoscan 201 UAVs was increased to 10-12 per team; total number of GNSS receivers was increased to 10;
а new version of autopilot firmware was introduced, which allowed increasing the number of flights of one UAV per day.

Additionally, the RTK mode of determining the coordinates of control points was applied (instead of the static mode); the number of control points per district was reduced, which allowed increasing measurement speed (from 20 CPs to 50 CPs per day, surveyed by one team) and shorten the time of geodetic field works.

More powerful chargers have been introduced.

The requirements for the resolution of digital images for the aerial survey of the territories between the settlements were changed from 7 to 9 cm/pix, which allowed to widen the boundaries of the surveyed area. The number of base flight stations at the takeoff point was reduced from two to one, which accelerated the preparation of the flights.

All these measures helped to significantly increase the productivity of field works and to develop the final technological scheme, which was applied for surveying the remaining districts of the Tula region conducted from March 23 to July 31, 2017.

First of all, the reconnaissance and examination of the points of the State Geodetic Network in the area of a particular district were carried out. Then a project for placing control points was created. The reference base stations were installed in such a way that the distance to the control points and the base flight stations would not exceed 30 km. A total of 24 RBS were installed, with help of which 4,000 hours of surveys were conducted. 2 to 5 teams consisting of 2 people were determining the coordinates of control points relative to RBS in RTK mode, using dual-frequency GLONASS/GPS receivers.

During the entire period of work, 21,000 hours of GNSS surveys were carried out at 90 points of the GGS and at more than 8,000 control points throughout the entire region with the determination of coordinates in MSK-71.1 coordinate system. Fig. 3 shows an example of placing of control points in one of the districts.

Fig. 3 – an example of the project of placing of control points in Yasnogorskiy district.

It should be noted that prior to the commencement of fields work in the test and other areas of the Tula region, the permissions of the General Staff of the Armed Forces of the Russian Federation, the authorities of the military region and the regional units of the Federal Security Services were obtained. Concurrently with the geodetic reconnaissance, UAV flights were being approved by the local administrations. Immediately before the start of the flights, the district administration was informed of the commencement of the works

During the flight planning, flight tasks were being drafted with the use of Geoscan Planner software (Figure 4), in which the area of the region was divided into flight zones. A base flight station was installed, relative to which the placement of photography centers was determined. The aerial survey was carried out simultaneously in different areas of the district by two teams of 4 people, each of whom controlled two Geoscan 201 UAVs (Geoscan 101 was not used during this period due to its lower productivity). After the landing of each UAV, the obtained digital images and data from the onboard GNSS receiver were uploaded on the operator’s computer. Then, the battery was replaced, new flight tasks were loaded into the autopilot, and UAV was launched again (in case of favorable weather conditions) to provide the continous functioning of UAVs until the very end of working day (Fig. 5). Thus, tens of thousands of images could be obtained per day. Such intensity of works allowed performing the survey of the districts in a very short period of time. For example, the aerial survey of Suvorovsky and Volovsky districts was carried out in 3 flight days, with the average time of performing the survey of an entire district being 9-14 days.

Fig. 4 – flight task in "Geoscan Planner" software interface.

Fig. 5 – a take-off of the plane.

The methodology of aerial photographing depended on the type of the surveyed territory and objects:

aerial survey of the settlements was conducted by Geoscan 201 UAV with a camera positioned perpendicularly to the horizontal plane and with the use of the oblique camera;
aerial survey of the territories between the settlements was conducted by Geoscan 201 UAV with a camera positioned perpendicularly to the horizontal plane;
aerial survey of the monuments and cultural objects was carried out with the use of Geoscan 401 quadcopter.

The required accuracy of orthomosaics and DTMs was achieved by abiding by a number of conditions: the transverse overlap of the images should exceed 50% (for settlements with the high density of high-rise buildings - 60%), and the longitudinal overlap of the images should exceed 70%. Furthermore, the spatial resolution for the settlements territory should not exceed 4 cm/pix, and for the territories between settlements - 9 cm/pix.

Every day, after the aerial survey was completed, flight field notes and geodetic field notes were compiled. Aerial photographs were sent to the military censor of the Western Military District. Further, aerial photographs and the data from GNSS receivers of the base flight stations and onboard GNSS receivers were transferred to the Geoscan headquarters, and later – to the remote sensing data processing department and the processing cluster located at the St. Petersburg Polytechnic University for the purpose of the geodetic equation and compilation of photographing centers catalogs.

Proper organization of field works allowed performing the survey in a short period of time, notwithstanding the vast area of the region. Thus, an aerial survey of 21 districts of the Tula region with an area of more than 25,000 km2 was conducted in 288 days, during which unmanned aerial systems stayed aloft for more than 10,000 hours.

Moreover, an aerial photographing of the architectural monuments of the Tula region with a resolution of 1-2 cm/pix was carried out by a separate team with the use of Geoscan 401 quadcopter. To achieve better accuracy and detail in the reconstruction of three-dimensional realistic models of objects, additional photos from the ground were captured in the zones inaccessible for the quadcopter. Based on the results of the data processing, 22 3D models of architectural monuments of the Tula region were constructed, among them - the Tula Kremlin, Kulikovo Field, the Yasnaya Polyana Museum-estate, and the Tula State Museum of Weapons (Fig. 6).

Fig. 6 - 3D model "Tula State Museum of Weapons".

Creation of orthomosaics and 3D models of the terrain and objects

Analysis and processing of the surveyed data for the purposes of creating orthomosaics and terrain models were carried out in the company’s office with the use of the software for automated photogrammetric processing - Agisoft PhotoScan Pro. The volume of images per one district usually amounted several hundreds of thousands (for example, during the survey of the Alexinsky district with an area of about 950 km2, 269 flights were conducted and 250,000 images were obtained), so this amount of data required colossal powers for the data processing and storage. To solve this problem, the company’s own computational cluster and the computational cluster of the supercomputer Polytechnic RSK Tornado (Fig. 7), the peak performance of which reaches 943 Tflops, were used. During the project, 300,000 Tflops of the calculations were performed with its use.

Fig. 7 - "Polytechnic RSC Tornado" supercomputer.

Data processing in Agisoft PhotoScan software implies the determination of the initial camera position and orientation (by the elements of the camera's external orientation), the definition of the coordinate system of the reconstructed model, and then the creation of a sparse point cloud on the basis of the images. To optimize the results of the determination of camera positions and parameters of their internal orientation, markers are placed and coordinates are specified. The markers are placed over the control points, allowing performing accuracy control, consisting of calculation of mean square errors of the control points with the known exact coordinates. If MSE does not exceed 10 cm (according to the requirements [4]), further processes are started.

Based on the calculated positions of the cameras, the program automatically calculates depth maps for each camera and creates a dense point cloud, and then - a three-dimensional polygonal model based on it, for which a texture is then constructed. For an orthomosaic creation basing on a dense point cloud or a polygonal model, a height field is constructed on which the orthophoto is projected.

As a result of processing of 6 million photographs, orthomosaics were created with a resolution of 5 cm/pix for the territories of settlements and 10 cm/pix for inter-settlement territories, as well as high-precision textured 3D models of settlements with 1,443 digital models of settlements. The volume of the obtained data was 218 terabytes.

Identification of violations of land legislation

Upon processing the results of the field measurements, the working group, which included about 25 GIS specialists, was detecting violations of land legislation basing on digital orthomosaics and information from the real property register (Ediniy Gosudarstvenniy Reestr Nedvizhimosti – EGRN) and cadastral plans of territory (CPT). Most specialists were engaged in vectorization of the actual boundaries of the land plots in order to identify violations in the field of cadastral registration with the use of Sputnik GIS and QGIS.

Besides, a geocoded address database was established through analyzing the data of the Federal Information Address System, address data of local municipalities and field surveys data. The number of the geocoded addresses of the database numbered 350 thousand.

The process of identification of violations of land legislation included:

vectorization of the boundaries of the actual land use of the land plots;
identification of previously registered land plots;
collation of the boundaries of the registered land plots, identification of cadastral registry errors;
identification of inappropriate utilization of agricultural lands.

For the organization of spatial data storage, an open PostgreSQL DBMS was selected that may be integrated into the QGIS environment and is suitable for the geospatial data operation. That allowed to create a smooth running infrastructure for the geospatial data operation, collection and storage.

Fig. 8 - list of vector layers in the interface of the Geoportal of the Tula region.

Based on the results of these works, a group of vector layers (Figure 8) was compiled. They contained information on:

land plots, information about which was available in CPT and which were registered by the state cadastral registration service (Gosudarstvenniy Cadastroviy Uchet - GCU);
land plots, information about which was also duplicated in the EGRN;
land plots, which had boundaries that intersected with the boundaries of sites, information about which was in the EGRN;
taxable land plots for which there was information about the tax payments of the landowners.

In addition, the following categories of land plots were distinguished for the settlements’ territories:

land plots which were registered in GCU (with determined boundaries), identified as a result of the orthomosaics interpretation by the boundaries of the actual land use;
previously registered land plots identified by the address and the area size from the CPT recordings and the geocoded address database information;
land plots which were not registered in GCU, identified as a result of the orthomosaics interpretation by the boundaries of the actual land use;
illegally used land plots (land grabbings), which were identified by comparing the area of the land plots registered in GCU with the area of the actual land use of such plots;
land plots, information about which was only available in the geocoded address database.

The agricultural lands layer was compiled according to the actual land use boundaries and the boundaries information from the EGRN, and split into categories - arable lands and non-arable lands, and then divided into subcategories, which allowed to determine the status of the land use:

for personal farming
arable lands
meadowlands
tree and shrubbery vegetation
perennial crops
inappropriately utilized agricultural lands

Overall, the following information about the land plots in the process of vectorization of the region was analyzed and revealed (Table 1):

Table 1. Analysis of the land use in Tula region
Category of the land plots	Number of the land plots	Total area, km2
Land grabbings	60 500	31
Appropriately utilized agricultural land plots	20 100	7 400
Land plots, overgrown by tree and shrubbery vegetation	9 500	1 700

Geoportal of the Tula region

For the creation, maintenance and visualization of all the obtained data at a convenient geoportal it was necessary to develop an effective network infrastructure.

As a result, a client-server web application was created with the use of Java programming languages (JEE), JavaScript, HTML, CSS. The server part organizes publication using the Open Geospatial Consortium (OGC) protocols. To provide a high load of the site server, duplication and geo-portal resource allocation on the base of the own servers (in a network storage of about 0.5 PB) are used, while the client part provides a user interface for the queries implementation and geospatial data visualization.

With regard to the structure of the geoportal, it is worth mentioning the following components with different functions:

Postgres/PostGIS DBMS - storage of geospatial data;
Web server nginx - transfer of the contents of web pages;
Tomcat 8 application server - container for JEE components, such as GeoServer, TerrainServer, TlsModelServer, which organizes publication of various types of geodata

The main solutions are implemented on the basis of GIS Sputnik Web. Work of the portal is optimized in such a way that 3D models do not require the loading of the user's processor, and the generation of the tile structure and tiles of orthomosaics occurs "on the fly."

The release version of the geoportal allows to view areas of the Tula region in three-dimensional space, either individually or as a catalog of models of distinct settlements (Fig. 9), enable all vector layers and their attributive information (Fig. 8), work with the portal database with the use of WMS protocol (for example, for adding the missing information), use measurement tools, change coordinate systems, etc.

Fig. 9 – display of a 3D model of a fragment of a settlement at the Geoportal of the Tula region.

Conclusions

Ultimately, all the necessary materials were obtained in the course of the work: DTMs, DEMs and orthomosaics, realistic 3D models of settlements and monuments. These materials meet the accuracy requirements for cadastral works and can be used for these purposes, as well as for touristic services, the organization of monitoring, territories management and for other purposes. The created Geoportal allows storing and updating spatial information and provides access to the resource for the local authorities and cadastral engineers.

As a result of performed works, violations of land legislation and errors of the cadaster and real estate registration were identified, zoning of territories was carried out and the potential economic impact of violations eliminating was determined. The potential economic impacts from the registration and taxation of the illegally used land plots (land grabbings) accounts for 3.4 bln roubles of one-time revenues and 165 mln roubles of annual revenues (according to NEO Center estimates), while the volume of the annual crop yields of the unused or inappropriately used agricultural lands amounts 5 bln roubles.

Finally, it is ought to be said that the technologies of Group of Companies Geoscan proved its high efficiency and high profitability. The completed project will give an impetus to the economiс and socio-cultural development of the region, as well as to the development of the program "Digital Economics of the Russian Federation".

Published in "Geoprofi" №2-2018 and №3-2018.

Reference list

1. High-precision cartographic 3D models of Singapore and Helsinki // Geoprofi. — 2017. — No 5. — P. 39-41.

2. 3D model of Tomsk city. — https://tomsk3da.admtomsk.ru/3d_city.

3. The Special Federal Programme «Development of a unified state system of registration of rights and cadastral registration of real estate (2014-2020). Approved by Resolution of the Government of the Russian Federation of 10.10.2013, No. 903 (amended by the Russian Federation Government Decree No. 1444 of December 22, 2016)».

4. Order No. 90 of the Ministry of Economic Development and Trade of the Russian Federation of 01.03.2016 «On approval of the requirements for accuracy and methods for determination of the coordinates of characteristic points of the boundaries of a land plot, requirements for accuracy and methods for determination of the coordinates of characteristic points of a building's contour, construction or an object of unfinished construction on a land plot, and requirements for determination of the area of the building, structure and rooms».

5. Federal Law No. 431-FZ of December 30, 2015 (amended on 03.07.2016) «On Geodesy, Cartography and Spatial Data and on Amending Certain Legislative Acts of the Russian Federation».