Multidimensional visualization analysis based on large-scale GNSS data

2.1 Hadoop

Hadoop is an integration of distributed storage and computing framework used for big data development. The core of Hadoop 2.0 includes the Hadoop Distributed File System (HDFS), Yet Another Resource Negotiator (YARN), and MapReduce modules (Tao et al. 2019, Sontakke and Dayanand 2019). The HDFS stores and manages massive data in the distributed file system. YARN is responsible for managing and scheduling cluster resources in master/slave mode. MapReduce is a distributed data processing model under Hadoop. An aspect of database selection, HBase (Bhupathiraju and Ravuri 2014) can query and obtain data at different granularities through the three query and access methods (row key, row key range, and full table scan).

2.2 Estimation of site coordinates

In the processing of navigation data, clock differences of the ground receiver and satellite may have an effect on calculating the distance between the two, as do the blur phases. Thus, we must estimate clock differences in a sequence of observations for each satellite and station, as well as estimating the blur parameter, which results in the rank deficit of the normal equation for estimating the station coordinates, clock differences, and blur phases simultaneously. The smaller the geometric network, the simpler it is to compensate for the dip shift of the network by adjusting the clock differences and blur phases. Therefore, external constraints are introduced into the equation that applies the nondifference or double-difference method. Usually, several reference coordinates are used to correct the estimating values of the station coordinates of the reference system. This correction is called the geodetic datum definition.

The geodetic datum definition used in the calculation of GNSS global coordinates is usually the least-constrained solution, which defines a seven-parameter transformation between a prior value [ x i , y i , z i ] T and an estimated value [ x ˜ i , y ˜ i , z ˜ i ] T of the station coordinates. The elements in the seven-parameter vector T = [ T x , T y , T z , α , β , γ , s ] T are three shifting parameters, three rotation parameters, and one scale parameter. The transformation is

and the matrix is

2.3 Visualization technology

Visualization is a mature technology nowadays. Determination of how to apply visualization technologies to analyze big data of navigation domains is the primary goal of this paper. The Baidu Map API (Wang et al. 2020) is an application interface based on the JavaScript language that supports HTML5. Its functions include map display, geographic analysis, region-based retrieval, and user-defined coverings. MapV is an open-source library for visualization of geographic information on a large scale. Cesium is an open-source platform for digital Earth developed by Analytical Graphics, Inc. (AGI) based on WebGL (Moser et al. 2010). The platform can render 2D, 3D, and 2.5D scenes of Earth without plug-ins, supporting a variety of data formats, e.g., GeoJSON and KML. Cesium repackages AJAX (Li and Peng 2012) functions to implement asynchronously requesting of large-scale geospatial data on the server side. Cesium also has proposed a visual language, CZML (Li and Yan 2021, Chaturvedi 2014), a subset of JSON, which uses key-value pairs to list the attributes and characteristics of objects such as the Earth or a satellite. In CZML, descriptions of different objects are separated by object id.

3.1 Big data platform design

A Hadoop distributed storage and computing platform was built to process GNSS data. The underlying hardware devices of the platform include clusters of distributed storage servers, cache servers, and computing servers. The HDFS and HBase distributed databases are used for data storage, and the MapReduce is used for computing framework for data processing. In addition, the platform utilizes Zookeeper as a cluster management tool, and application development languages include Java, Python, Perl, and Fortran, among others.

GNSS data include downlink broadcast messages, observation data, meteorological data, time-difference measurement data, interference detection, and multipath detection data of the four major navigation satellite systems. As shown in Figure 2, the data come from global observation stations. The observation data and broadcast messages are stored in RINEX format files, and the file contents must be analyzed first. According to the time interval, the data are divided into day, hour, and 15-min ultra-fast data. In addition to eliminating invalid data, data preprocessing must splice the continuous measurement data within a day, filter out the data in overlapping periods, and make data corrections on a daily basis. Observation data and broadcast messages are stored in HDFS and HBase distributed database, respectively, after preliminary governance and statistics are completed. Satellite orbit files are generated after further calculation and analysis. The statistical results are finally presented in two dimensions in the visual system, and the satellite orbit is presented in three dimensions.

Figure 2

Data-processing architecture.

The structure of the satellite ephemeris is presented in Table 1; the row key is a string of date and time, and the value is the coordinate value under the Earth-Fixed (Guo et al. 2021) system. The structure of the navigation data table is shown in Table 2; the row key is the station name and date. The data may be unusable due to incomplete data, satellite failure, or other reasons. In the present work, for example, both received observations (R:R09=60 in the table) and valid observations (R:R09_v=0 in the table) were analyzed.

3.2 Data visualization

The proposed GNSS data visualization system visualized GNSS data stored in the distributed storage system with the 2D map module and 3D space module. The 2D module shows the ground system including the distribution and status of station networks, while the 3D module shows the status and subsatellite points of the satellite system. The system adopts a three-layer architecture, i.e., data access, business logic, and presentation layer. The data access layer is used to interact with distributed storage systems, while the presentation layer provides a user interface. The business logic layer implements the logic operations of data, such as data merging or filtering.

4.1 System environment

The system environment mainly includes the software and hardware environments of the visual platform development, Hadoop cluster, and GNSS data database. The details are presented in Tables 3 and 4.

4.2 System implementation

4.2.1 GNSS measurement network module display

Based on MapV, the homepage of the station network dynamically displays the data flow diagram from iGMAS stations into the data center. As shown in Figure 3, the flow graph includes iGMAS stations, the data center, and traces between them.

Figure 3

Flow graph.

The names of the four major navigation systems – GPS, GLONASS, GALILEO, and BeiDou – are labeled on the top of the map. The user selects the system to view the information of all stations that track the system. AJAX is used to refresh the view asynchronously, and the data are updated locally without refreshing the entire page. The data quantities of all stations are dynamically represented by a heat map, with red representing a large quantity of data and blue a small quantity. The darker the color, the larger the volume of data. When the mouse hovers over a station’s position, the name, coordinates, and some statistics of the station are displayed (Figures 4 and 5).

Figure 4

Stations contained in the GPS system.

Figure 5

Retrieval.

By adding a click-trigger event to the map layer, clicking a station’s position will display the station’s statistics page. As shown in Figure 6, the statistics page displays more working information for the selected station. Because the data collected by the station may contain some errors, such as unmatched dates, which make the data unusable, the usable and collected data are counted separately. The page organizes those statistics by days for comparison so that users can conveniently understand how the station contributes to the entire navigation system. This information is useful in helping to filter the data while performing orbit calculations, or in planning the navigation missions.

Figure 6

GRAS station statistics page.

4.2.2 3D visualization display

The layout of the display interface is divided into four parts: virtual Earth, function module, plug-in functions, and timeline display module. The virtual Earth is in the middle area, which is also the area for 3D scene display and function-module realization. Users can perform basic interactions, such as mouse flipping and zooming, based on the virtual Earth. The function module in the top area of the interface includes global options, satellite control, target center perspective, and equatorial perspective modules (Lian et al. 2021). The menus are all multilevel menus. The related plug-in functions based on the Cesium, including perspective replacement, and 2D and 3D switching, appear on the right-hand side of the display interface. The timeline display module is at the bottom of the interface. The overall layout of the interface is designed to put the functional module bar on the side of the page, around the visual interaction page in the middle, so that the area reserved for the visual scene display space is greater (Figure 7).

Figure 7

GPS satellite constellation.

Choosing and clicking a satellite on the main interface will show the satellite control panel on the left-hand side of the interface. Taking the BeiDou C08 satellite as an example, as shown in Figures 8 and 9, the satellite control panel lists all satellite operations. Clicking the “Track” option can draw the trajectory in the inertial coordinate system or the Earth-Fixed coordinates system. Cesium provides the computeIcrfToFixedMatrix() function to convert Inertial coordinates into Earth-Fixed coordinates, in which the parameter of time is set in an asynchronous manner. The SampledPositionProperty() function generates a pair of time and position that is set to the “Entity” function as the “position” property to generate an entity, which also includes properties such as “Path,” “Label,” “Ellipsoid,” “Billboard,” and “Model.” Dragging the “Trail Time” and “Lead Time” axes can set the time to show the trajectory animation, which changes the “Path” attribute. The “Track Lines” option selects the track as a dashed or solid line. Dragging the “Red,” “Green,” and “Blue” option axes can set the RGB values of the track color. The time axis is displayed at the bottom of the interface. Users can view the simulation animation of the satellite’s movement within a certain period.

Figure 8

Satellite control panel.

Figure 9

Earth-fixed coordinate system.