GPR surveying method as a tool for geodetic verification of GESUT database of utilities in the light of BSI PAS128

In order to verify the accuracy of locating utilities using the above-mentioned solutions, a study of three combinations of the systems were designed (Fig. 3). The studies were performed at the AGH test base at Mydlniki (Cracow), with 9 pipes of various diameters, made of different materials and embedded there at different depths (Ortyl, 2006). Their coordinates were captured by direct methods, constituting a proper reference for the results of indirect methods. The test base is located in a difficult terrain, overgrown with dense grasses in the summer, with a slope of approximately 4.5%. The local control points were measured using the Leica GS16 receiver in the RTN surveying mode. Using the method of resection, the position of the total station was tied to these points with the accuracy of ±0.01 m. Based on the known profile of the pipes, two transverse profiles were designed, set out and marked, along which the antennas of the GPRs were directed (Fig. 2).

Figure 2

Figure 3

The first step involved the performance of a survey using the RAMAC/CU II GPR system with the antenna with the frequency of 500 MHz. The measurement of a single trace was triggered at the distance intervals of 0.03 m. In order to improve the S/N (signal to noise) ratio, the number of stacks equal to 32 and the sampling rate 10 times greater than the dominant frequency of the antenna was assumed. The time window with a length of 96 ns at a wave speed of 0.01 m/ns allowed the penetration of the medium to a depth of approximately 5 m. The Leica Nova MS50 total station and the 360 prism mounted on the dedicated adapter, 0.47 m above the ground, were used as a system for locating the position. The transmission of the coordinates in the 2000 system was carried out with Satteline 3ASd radio modems, using the PseudoNMEA protocol. Eight profiles were performed – two passes in opposite directions for both determined profiles.

Another combination included the compact DS2000 GPR system with two antennas with frequencies of 250 MHz and 700 MHz. The same coordinate transfer system was adopted, as well as the locating instrument and the measurement step equal to 0.03 m. The length of the time window was extended to 120 ns. The available set did not have a dedicated adapter for the surveying instrument, therefore, the 360 prism on a two-metre-high pole was attached to the handle of the trolley. One pass for both profiles was performed.

The last proposal for tying echograms to georeferencing was locating the coordinates for the DS2000 GPR system by the GNSS system in the RTN surveying mode. This method differed from the integration with the total station mainly with respect to the format and the way of data transmission. This was due to the fact that in this solution, thanks to the Bluetooth communication, the coordinates reached the GPR in the

WGS84 system converted to the UTM 34N system.

In this study, the echograms from the three systems were subjected to a filtration sequence in the ReflexW programme. The signal was normalised, the background was removed and the bandpass frequency filters and the gain filter were used, which facilitated the identification and vectorisation of the pipes along the signals with the highest amplitude, forming a shape similar to a hyperbole. The results were exported to ASCII files where, according to the criterion of the shortest time of the beam pass, the numbers of the traces and the relative depths of the samples corresponding to the highest point of the pipe were determined.

In order to locate the measurement results from the Mala GPR system, the coordinates in the 2000 system were used, saved by the acquisition programme in the *.cor file. This allowed for simple import operations, the interpolation of the duplicated data and the integration of the echograms with the coordinates in the ReflexW programme (Fig. 4b). Although this programme has the ability to import data in the format appropriate for the IDS systems, the interpretation of an implicit georeferenced file caused the occurrence of the translation and scaling of the coordinates and, consequently, errors in the traces exceeding 10 metres. Therefore, the uNext programme was used to tie the coordinates with the results of the surveys of the DS2000 GPR system (Fig. 4a). For that data, the analysis of the distance of the pipeline determined by the GPR method relative to the sections representing the actual profile of the pipes was performed, as well as the analysis of the compliance of their embedment depths.

Figure 4

Another survey was carried out in Cracow at Reymonta street, in the grassy area at the Faculty of Animal Breeding and Biology at the University of Agriculture. The choice of this specific area was dictated by its accessibility (without the need to stop road traffic) as well as the multiplicity and diversity of utilities, located parallel to each other, at distances not exceeding 3.5 m. Based on the GESUT data obtained from the County Geodetic and Cartographic Documentation Centre, an area with 8 utility lines was selected (Fig. 5): sewage pipe (koD300), low pressure gas pipe (gnD300), medium pressure gas pipe (gs90), 4 medium voltage power cables (eSA₁, eSA₂, eSA₃, eSA₄) and a bundle of 3 medium voltage power cables (3eSD). A grid of 26 profiles with a mesh of 1 m × 1 m constituted the base for the surveys (Fig. 5). The longitudinal profiles of the grid were located parallel to the axis of the utility lines.

Figure 5

During the survey performed with the RAMAC/GPR CU II GPR system with a 500 Hz antenna, as in the case of the experimental research, the positioning was conducted with the MS50 total station, because the data of this solution was analysed for compliance with the available documentation. The DS2000 GPR was also used for the surveys, but without the possibility of real-time data positioning, so they were used only for the visual analysis of the echogram quality.

In order to determine the compatibility of the GPR data with the existing cartographic study, first of all, it was necessary to capture information on the profile of the utilities based on the acquired echograms. In order to facilitate the initial process of utility network identification, the GESUT documentation was used, based on which the approximate distances of the subsequent utilities were determined from the origins of the measurement profiles. The precise positions of the points representing the axes of the pipes on the echograms were captured based on the vectorisation of the signal with the highest amplitude and the identification of their tops according to the shortest time criterion. Finally, the profile of the straight line segments of the networks was determined using linear regression with the minimised observation distance from the straight line (Fig. 6). The adjustment was made for the two variants of determining the coordinates of the pipe tops:

Figure 6

The distances between the approximate profiles of the utilities and the available documentation have formed the basis for the planar verification of the network compliance. The in-depth profile of the utilities along the test sections was also analysed.

The last stage of the study involved a visual comparison of the quality of the data captured with the use of the 500 MHz Mala antenna, as well as the 250 MHz and 700 MHz Leica antennas, for the effectiveness of utility detection.

Before presenting the numerical data illustrating the accuracy of the systems during the performance of the experiment in Mydlniki, it should be noted that the echograms of the DS2000 GPR system are characterised by a better quality and contrast of images, which has allowed to identify the pipes with 50% efficiency for both profiles. In the case of the Mala system, out of the expected 36 observations for a single profile, 15 and 7 pipe tops were reliably identified for profile 1 and profile 2, respectively. None of the echograms, regardless of the system used, identified the pipes R3, R4 or R6.

The mean square position error of the pipe tops identified on the RAMAC/CU II GPR echograms relative to their base profile was ±0.10 m. This error largely results from the necessity of interpolating the coordinates in places where there was a delay in radio data transmission or because the acquisition programme duplicated the captured coordinates for several subsequent traces. For the DS2000/TS and DS2000/GNSS combinations, this error assumed the values of ±0.23 m and ±0.24 m, respectively. The errors induced by the Leica system significantly exceed the error obtained by the Mala system results. Thus, the errors of determining the position of this GPR in the direction transverse to the measurement profiles were also determined, and they reached ±0.29 m and ±0.19 m. Such distribution of errors, both relative to the pipes and to the profiles, may point to the influence of a slope on determining the position of a high-placed target which, with the height of approximately 2.20 m above the ground and a slope of 4.5%, induces an error of ±0.01 m.

In order to ensure that such conclusions were right, another test was carried out to supplement the experiment, in which the coordinates of the baselines located before and on the slope of the hill with the slope of 45% were determined in the RTK-RTN measurement mode (Fig. 7)

Figure 7

When approaching the baselines with the GPR, the coordinates were read out in the uNext programme, however, this time the data was also processed using the available dedicated software. GRED HD is a powerful GPR data processing tool created by the IDS GeoRadar brand. The programme enables the creation of advanced 3D data analysis and visualisation as well as conversion of vectorisation to CAD formats. Leica has implemented similar solutions in its own DX Office Vision software, dedicated to the detection and digitalisation of utility lines based on the DS2000 GPR surveys.

The coordinates captured directly from the acquisition software were located at a distance of over 1.08 m from the baseline 2, which proves that no pole tilt adjustment was included in the programme. The processing of the data in the GRED HD programme improved the results, but with such a large slope of the terrain, differences of 0.36 m were obtained from the actual position of the baseline. The DX Office Vision programme proved to be the best tool in terms of the accuracy of integration of surveying data and echograms. The baseline coordinates obtained in it were located at a distance of 0.15 m from the actual profile of the line.

The parameter of the in-depth location of utilities is as important as their horizontal position for designing investments associated with the area where these utilities are located. During the analyses, there was a high repeatability of the results for each used systems. For the Mala system, there was a 78% total consistency of the depth of the pipe tops read out from the echograms for at least 2, and in some cases even 4 different passes; for the others, the differences did not exceed the value of 0.01 m. For DS2000, the discrepancies between the two passes fell within 0.01 m–0.03 m. Despite such high internal accuracy of both systems, the differences between them reached even 0.06 m and the deeper the pipe was embedded, the greater the difference noted. It is worth emphasising, however, that for the Mala system, having taken into account the terrain elevation, the differences in the value of the Z-coordinate of the pipe tops did not exceed 0.02 m.

The knowledge of the exact material characteristics of the pipes embedded in the test field allowed us to confirm the dependence of the polarisation of the wave reflected from the material of the pipes. The Mala system recorded positive values of amplitudes of the signal reflected from the steel pipes, and negative values for the PVC pipes (Radzevicius and Daniels, 2000). The reverse polarity was demonstrated by the signal emitted by the antennas of the DS2000 system (Fig. 8).

Figure 8

Verification of the results of the GPR surveys in relation to the GESUT cartographic data should have been started with a visual evaluation of the echograms. This is a preliminary, but very important part of the analysis, because the high amplitude signals are identified and classified at this stage. The data captured at Reymonta street turned out to be extremely difficult to interpret. Under favourable conditions of a geological substrate, the signal reflected from the upper surface of the linear utility should have the shape of a hyperbole. In the case of RAMAC/GPR CU II echograms, on the traces transverse to the utility lines, uniformly interpretable signals come only from the bundle of the power lines (3eSD), the gas pipe (gnD300) and the unidentified utility line located below the 3eSD bundle. For the power cables (eSA₂, eSA₃, eSA₄) there was a strong, elongated signal resulting from mutual interference of the reflected signals, so it was impossible for individual components to be visually distinguished. The power line (eSA₁) with the profile coincident with the gas pipeline (gs90) within the profiles 9–13 (Fig. 9) was not identified.

Figure 9

It could therefore be supposed that this power cable does not really exist. The analysis of the longitudinal profiles (Fig. 10) revealed the occurrence of a fragment of this utility line after changing its profile from the meridian to the latitudinal one, so this hypothesis was rejected in favour of another one that the gas pipe probably covers this power cable.

Figure 10

For the other utility lines, that is, koD300 and gs90, there was no signal unambiguous for interpretation, so the fragments of the echograms were distinguished based on the documentation. The in-depth analysis of the signals with a strong response in the places where the sewage pipe was likely to occur, has demonstrated rapid changes in the depth of its embedding, reaching 0.60 m on a one-metre-long section. In the execution of pipelines, especially general sewage systems, no slopes of 60% were applied. Therefore, this pipeline profile was considered incorrect, and the koD300 sewage pipe impossible to be detected only based on the available data.

The coordinates of the individual traces are encumbered with the error of their positioning method, which results in obtaining a broken axis of the utility line based on the vectorisation of the echograms, which forces the approximation of the shape of the axis to a more real straight line. Table 2 contains the values of the regression line fitting errors for the coordinates determined by the MS50 total station and for the coordinates interpolated along the profiles. Table 3 illustrates the values of the distances of the determined profiles relative to the base map. The method of positioning data by interpolation of the coordinates is less accurate (reveals the influence of the calibration of the encoder measuring the distance from the baseline of the profile origin), which is illustrated not only by the tabular data but also the visual consistency of the profiles. The limit values obtained for both methods do not exceed 0.30min most cases. The exception is the gas pipe gnD300, for which the offset of the measurement data from the base data is about 0.50 m. This discrepancy is of a constant nature over the entire length of the measurement; so, it can be inferred that the position of this pipeline in the GESUT database is incorrect.

The GPR survey carried out with the use of the Mala system has confirmed the location of the utility lines: eSA₂, eSA₃, eSA₄, 3eSD and gs90; it has indicated the probable correct profile of the gnD300 pipeline; it has detected an unidentified power cable located below 3eSD; it has confirmed the existence of eSA₁; and it has not confirmed koD300.

Contrary to the expectations (Reichel, 2017), the data captured by the DS2000 GPR at Reymonta street were characterised by slightly lower quality. On the echograms of the 700 MHz antenna, there were no signals of reflection from the pipelines koD300, gs90, gnD300 or the alleged utility line identified, thanks to the RAMAC/CU II GPR. In the part of the echogram that is near the surface, the resolution and contrast were satisfactory, however, below the depth of 1 m, the data were strongly noisy and the useful signals could not be distinguished. The echograms of the 200 MHz antenna were a good complement to the data at lower depths (Fig. 12).

Figure 11

Figure 12

Capturing the profile of the utility network based on the vectorisation of echograms on rectilinear segments has yielded very good accuracy results. However, this is a multi-stage and time-consuming process. The longer the length of the utility line, the greater the effort required to perform repetitive activities, that is, analysis and vectorisation of subsequent echograms – B-scan. Only after the data has been processed for the entire segment, is it possible to perform the approximation of the axis profile. According to the authors of this publication, for the needs of the GESUT database, this process can be replaced by the vectorisation of the profile of the utilities identifiable on the horizontal slices of the C-scan of the 3D model performed at the appropriate depths (Fig. 13).

Figure 13

The quality of 2D data processing into a three-dimensional model depends largely on the selected interpolation parameters (Lu et al., 2018) and the adopted coordinate system. There are two main ways of positioning echograms in space. The processes and problems described here are adequate for the ReflexW software, in which the processing has been carried out.

The first method involves the creation of a 3D model in a local system, for which the point of origin and the X axis are determined by the first profile of the model. The origins of subsequent echograms are arranged in accordance with the assumed profiling step along the Y axis. The interpolation is performed along the axis of the adopted system. This method is most effective in the event of the maximum fulfilment of conditions such as: the parallelism and equidistant increment of the subsequent profiles, the measurement step and the number of traces in the profile. The interpolated data fills tightly and closes within the boundaries of the studied area (Fig. 13).

The second method is to use the georeference for positioning echograms. The interpolation is performed along the main axes of the global system in which the GPR data are embedded. The selection of the interpolation parameters was carried out based on the tests of subsequent variants, since the acceptance of the values recommended by the manufacturer, based on the relative distribution of the profiles and the measurement step, would result in capturing data with significant deficiencies in the model continuity (Fig. 14). In addition, the programme was extrapolated on the edges of the area.

Figure 14

The model interpolated based on echograms and their coordinates can be arbitrarily cut along its main axes, also generating horizontal sections with a given depth and thickness. Strong signal refiections can be emphasised by further filtrating transformations and they can be exported to the geoTIFF form. It should be remembered that the ReflexW programme operates in a mathematical system, and therefore, it is necessary to calibrate it (regardless of the positioning method chosen) in the programmes that operate on surveying systems.

Most CAD programmes have the ability to load *.TIFF files and use them as references to create cartographic studies regarding, for example, vectorisation of the axes of utility lines (Fig. 15). The differences obtained between the axes and the regression lines, based on the coordinates of the subsequent tops of pipes in the studied area do not exceed ±0.30 m.

Figure 15