**Assessment of DSMs Using Backpack-Mounted Systems and Drone Techniques to Characterise Ancient Underground Cellars in the Duero Basin (Spain)**

**Serafín López-Cuervo Medina 1,\*, Enrique Pérez-Martín <sup>2</sup> , Tomás R. Herrero Tejedor <sup>2</sup> , Juan F. Prieto <sup>1</sup> , Jesús Velasco <sup>1</sup> , Miguel Ángel Conejo Martín <sup>2</sup> , Alejandra Ezquerra-Canalejo <sup>3</sup> and Julián Aguirre de Mata <sup>1</sup>**


Received: 30 October 2019; Accepted: 2 December 2019; Published: 4 December 2019

**Abstract:** In this study, a backpack-mounted 3D mobile scanning system and a fixed-wing drone (UAV) have been used to register terrain data on the same space. The study area is part of the ancient underground cellars in the Duero Basin. The aim of this work is to characterise the state of the roofs of these wine cellars by obtaining digital surface models (DSM) using the previously mentioned systems to detect any possible cases of collapse, using four geomatic products obtained with these systems. The results obtained from the process offer sufficient quality to generate valid DSMs in the study area or in a similar area. One limitation of the DSMs generated by backpack MMS is that the outcome depends on the distance of the points to the axis of the track and on the irregularities in the terrain. Specific parameters have been studied, such as the measuring distance from the scanning point in the laser scanner, the angle of incidence with regard to the ground, the surface vegetation, and any irregularities in the terrain. The registration speed and the high definition of the terrain offered by these systems produce a model that can be used to select the correct conservation priorities for this unique space.

**Keywords:** DSM assessment; backpack mobile mapping; UAV; underground cellars

### **1. Introduction**

Instruments and techniques for the massive capture of data are increasingly being used to document all types of cultural landscapes and heritages. There are recommendations and criteria for adequate procedures to ensure the defence and preservation of these types of heritages and landscapes in the European context [1]. Many research projects on heritage management use geospatial technologies [2–5] to generate various products such as digital surface models (DSM) [6–13]. A large number of studies focus on optimising and improving actions with unmanned aircraft vehicles (UAV) [14].

Against this background, our study zone, the underground cellars of El Plantío in Atauta (Soria), declared an Asset of Cultural Interest (Bien de Interés Cultural, BIC) in March 2017, which represents a unique testimony of the life associated with work on the land and the traditional wine production system. It is important to ensure that they are not ultimately forgotten and destroyed, as they are a manifestation of the cultural identity of a large region in the Duero River basin (Figure 1).

In recent years, fixed terrestrial laser scanning (TLS) systems have been used to map and monitor various areas of interest from the point of view of their cultural heritage [15–18], in archaeological studies [19,20], underground studies [21,22], and civil engineering [23], among others [24]. 3D technologies include both the TLS system and sensors installed on UAV for the documentation, visualisation, and preservation of heritage [25–27]. The precision of a digital surface model (DSM) obtained by UAV photogrammetry for documenting surface structures [28] and forms in 3D models [29–31] and for identifying constructions in urban planning [32,33] has also been analysed. Chiabrando et al. [19] described a technique for preparing digital surface models in their archaeological studies. Norhafizi validated the use of UAV for creating DSMs of tide data [34]. Villanueva [35] studied DSMs and their application in zones at risk of flooding.

Komarek et al. [36] carried out studies to assess the precision of DSMs obtained by UAV on rural plots. In other cases, DSMs have been assessed with or without ground control points (GCP) taken by means of a real-time kinematic Global Positioning System (RTK-GPS) techniques [37–39]. Other studies have evaluated the use of software for generating DSM models from UAV with GCP taken in the field by means of other geomatic methods such as TLS, GPS [ *Sensors* **2019** 40], or UAV photogrammetry techniques [41]. , *19*, x FOR PEER REVIEW 3 of 20

**Figure 1.** Location of the study area. The underground cellars are located in Atauta (Soria), a region in the Duero Basin in the north-central part of Spain. **Figure 1.** Location of the study area. The underground cellars are located in Atauta (Soria), a region in the Duero Basin in the north-central part of Spain.

*2.2. Equipment Used to Take the Images*  The topographic and photogrammetric survey using UAV techniques was carried out with a Mavinci fixed-wing drone with a Lumix-GX1 camera with a focal distance of 14 mm and a resolution of 4592 × 3448 pixels. The flight was made at a height of 90 m, obtaining a ground sample distance There are examples of studies on the precision of 3D models in indoor spaces and areas through backpack mobile mapping [42–44] and assessments of different TLS [45–47]. Other works assess the performance of a mobile mapping system (MMS) and backpack mobile mapping [48,49] compared to the use of a TLS [47], and highlight the superior performance of the mobile system, even though its overall precision is lower than with TLS. Campos [50] applied the backpack mobile mapping system

techniques, high-precision control points were observed in order to improve the geo-referencing of the photogrammetric model with points on the ground and for a quality check assessment. The deformation in the images caused by the camera [54] is compensated with an autocalibration

**Figure 2.** (**a**) Mavinci UAV used during the registration process. (**b**) Image orientation process task in Agisoft Photoscan® professional software (http://www.agisoft.com/) showing flight paths and

performed during the aerotriangulation process (Figure 2).

photocenters. (**c**) 3D view diagram of the recorded images.

(GSD) of 2.16 cm and taking 344 images by following a flight plan with a regular grid pattern defined

in forest mapping where there is limited accessibility. Przemyslaw [51] also used backpack mobile mapping for the geolocation of tree trunks and their comparison with UAV records.

The DSMs obtained are used to compare and complement the information registered with ground penetrating geo-radar technology (GPR) and terrestrial laser scanning (TLS) [52]. In previous studies in the same area [53], DSMs enabled the definition and comparison of wall and ceiling thicknesses in underground cellars to ensure their stability.

The aim of this work is to assess the condition of the roofs of underground cellars in their natural state by obtaining an accurate DSM to characterise, detect, and prevent their collapse. This is accomplished using a backpack-mounted 3D mobile scanning system (backpack MMS) and a fixed-wing drone (UAV).

The study examines the use of backpack MMS in areas of irregular terrain such as the Atauta underground cellar, using data registration parameters and DSM generation including working widths and angles of incidence in the measurement and obstacles. The study also compared it with UAV equipment that enables analysis of the benefits of both systems for generating DSMs of key importance in assessing the current risk status of underground wine cellars.

#### **2. Materials and Methods**

#### *2.1. Case Study Description*

The study area is part of the series of underground wine cellars of El Plantío in Atauta, Soria (41◦310 N, 3◦120 W), shown in Figure 1, which were declared an Asset of Cultural Interest (BIC) as an "ethnographic collection" on 16 March 2017. Located at the foot of the village of Atauta on an area of 1.9 ha, they are a testimony of the life associated with work on the land and the artisanal wine production system. These underground constructions were used to store and preserve wine.

#### *2.2. Equipment Used to Take the Images*

The topographic and photogrammetric survey using UAV techniques was carried out with a Mavinci fixed-wing drone with a Lumix-GX1 camera with a focal distance of 14 mm and a resolution of 4592 × 3448 pixels. The flight was made at a height of 90 m, obtaining a ground sample distance (GSD) of 2.16 cm and taking 344 images by following a flight plan with a regular grid pattern defined in a northwest-southeast direction. The longitudinal overlap was 70%, and the transversal overlap was 50%. Precise coordinates were obtained throughout the flight due to an RTK-GPS receiver located on the UAV, which received corrections from a base on the ground. Also using RTK-GPS techniques, high-precision control points were observed in order to improve the geo-referencing of the photogrammetric model with points on the ground and for a quality check assessment. The deformation in the images caused by the camera [54] is compensated with an autocalibration performed during the aerotriangulation process (Figure 2).

#### *2.3. System Used for Mass Registration with a Laser*

A Leica Pegasus data registration system was used as a backpack mobile mapping model (Figure 3). This system incorporates five cameras and two LiDAR (light detection and ranging) scanners for the registration of the 3D point cloud and images. It also has two Velodyne sensors (VLP16) that rotate at 10 Hz and acquire 600,000 points per second at a distance of up to 100 m, even though this may be influenced by several factors, as indicated in its specifications [48,55]. It weighs approximately 13 kg and has a scanning autonomy of three hours. The system includes a SLAM (simultaneous localisation and mapping) algorithm and an IMU (inertial measurement unit) as an aid for generating the 3D model.

**Figure 1.** Location of the study area. The underground cellars are located in Atauta (Soria), a region

The topographic and photogrammetric survey using UAV techniques was carried out with a Mavinci fixed-wing drone with a Lumix-GX1 camera with a focal distance of 14 mm and a resolution of 4592 × 3448 pixels. The flight was made at a height of 90 m, obtaining a ground sample distance (GSD) of 2.16 cm and taking 344 images by following a flight plan with a regular grid pattern defined in a northwest-southeast direction. The longitudinal overlap was 70%, and the transversal overlap was 50%. Precise coordinates were obtained throughout the flight due to an RTK-GPS receiver located on the UAV, which received corrections from a base on the ground. Also using RTK-GPS techniques, high-precision control points were observed in order to improve the geo-referencing of the photogrammetric model with points on the ground and for a quality check assessment. The

*Sensors* **2019**, *19*, x FOR PEER REVIEW 4 of 20

3). This system incorporates five cameras and two LiDAR (light detection and ranging) scanners for

in the Duero Basin in the north-central part of Spain.

*2.3. System Used for Mass Registration with a Laser* 

*2.2. Equipment Used to Take the Images* 

**Figure 2.** (**a**) Mavinci UAV used during the registration process. (**b**) Image orientation process task in Agisoft Photoscan® professional software (http://www.agisoft.com/) showing flight paths and photocenters. (**c**) 3D view diagram of the recorded images. **Figure 2.** (**a**) Mavinci UAV used during the registration process. (**b**) Image orientation process task in Agisoft Photoscan® professional software (http://www.agisoft.com/) showing flight paths and photocenters. (**c**) 3D view diagram of the recorded images. in conditions of low light. The work software has tools for extracting LiDAR and photogrammetric data and detecting changes, and is compatible with workflows by means of AutoCAD and ArcGIS. The only limitation to its use is its autonomy of four hours due to its batteries.

**Figure 3.** (**a**) GPS Leica GX1230 GG equipment, acting as a reference base. (**b**) Pegasus backpack system for taking the point cloud in the study area before obtaining the DSM model to be assessed. **Figure 3.** (**a**) GPS Leica GX1230 GG equipment, acting as a reference base. (**b**) Pegasus backpack system for taking the point cloud in the study area before obtaining the DSM model to be assessed.

It took 30 min to configure and calibrate the backpack MMS data registration system, and one hour to survey the study area. The main features of the UAV and the Pegasus backpack are shown in Table 1. **Table 1.** UAV and MMS system specifications. **Main Features Mavinci UAV Pegasus Backpack**  Technology Lumix-G1 16 Mpx camera Velodyne VLP16 laser scanner Measurement technology Computation from images Polar distance measurement Leica Pegasus Backpack allows the acquisition of LiDAR and image data with precise positioning of outdoor and indoor data [18] in a system that is easily transportable by one person. This makes this type of equipment very useful in environments with limited space, underground environments, [20] and in areas with dense vegetation, as well as for managing data on disasters [56] and documenting industrial facilities. It offers the option of including external sensors such as GPR equipment, thermal cameras, noise and pollution sensor, etc., assisted by a flash module for working in conditions of low light. The work software has tools for extracting LiDAR and photogrammetric data and detecting changes, and is compatible with workflows by means of AutoCAD and ArcGIS. The only limitation to its use is its autonomy of four hours due to its batteries.

Distance measurement 90 m 5–100 m System resolution GSD 2.16 cm Dist. acc. 3 cm at 100 m DSM resolution 600 points/m2 36,000 points/m2 It took 30 min to configure and calibrate the backpack MMS data registration system, and one hour to survey the study area. The main features of the UAV and the Pegasus backpack are shown in Table 1.


**Table 1.** UAV and MMS system specifications.

backpack, as shown in Figure 4.

#### *2.4. Control and Assessment System Obtained by GPS* GCPs and GEPs observed were located at a radius of less than 200 m and measured with another two GPS rover receivers. From this base, 12 GCPs were computed in post-processing to support the UAV.

High-precision coordinates were computed on a fixed base using GPS techniques in post-processing to obtain both Ground Evaluation Points (GEP) and Ground Control Points (GCP). The points are distributed to cover the entire area of characterisation and possible collapses on the surface of the underground cellars, which are the same as the UAV flight and the tracks taken by the backpack, as shown in Figure 4. The GCPs were positioned on the periphery and inside the study zone. This same base served as a support for the backpack MMS, and all its tracks were within a range of less than 250 m from the base. Furthermore, 59 GEP coordinates were also computed by RTK-GPS techniques from the same base, and, subsequently, used in the assessment of the DSMs generated.

*Sensors* **2019**, *19*, x FOR PEER REVIEW 5 of 20

of the underground cellars, which are the same as the UAV flight and the tracks taken by the

0.02 m in all the points. The base GPS receiver was situated in the centre of the study area and all the

This GPS base served as a reference for all the topographic and photogrammetric surveys (see Figure 5, Section 1, centre). The equipment used was a Leica Geosystems GX1230 GG, and the data

**Figure 4.** GEP observed using RTK-GPS techniques for the characterisation surveys of collapse zones. Lines depict the backpack mobile mapping tracks (Tracks A to F) used to assess the 3D model. **Figure 4.** GEP observed using RTK-GPS techniques for the characterisation surveys of collapse zones. Lines depict the backpack mobile mapping tracks (Tracks A to F) used to assess the 3D model.

*2.5. Methods*  As stated, the aim of this research is to obtain digital surface models (DSM) on the terrain corresponding to the natural cover of the underground cellars. A methodology was defined in several phases for this purpose, as shown in the diagram in Figure 5. In the first phase, the 45 GEP points were analysed after obtaining the data with methods based on backpack MMS and UAV technologies. These points were identified using Agisoft Photoscan® Professional software and included in the aerotriangulation process, and then identified and extracted from the DSM generated from the dense This GPS base served as a reference for all the topographic and photogrammetric surveys (see Figure 5, Section 1, centre). The equipment used was a Leica Geosystems GX1230 GG, and the data were processed with the Leica Geo Office software, with a relative average precision of more than 0.02 m in all the points. The base GPS receiver was situated in the centre of the study area and all the GCPs and GEPs observed were located at a radius of less than 200 m and measured with another two GPS rover receivers. From this base, 12 GCPs were computed in post-processing to support the UAV. The GCPs were positioned on the periphery and inside the study zone. This same base served as a support for the backpack MMS, and all its tracks were within a range of less than 250 m from the base.

point cloud in the UAV photogrammetric project. Each GEP point was located in the data from the photographs registered by backpack MMS and measured by stereoscopy with ArcExplorer (Esri, USA) software. The GEP were also identified in the LiDAR point cloud from the backpack MMS. These four Furthermore, 59 GEP coordinates were also computed by RTK-GPS techniques from the same base, and, subsequently, used in the assessment of the DSMs generated.

measures were compared with the GEP coordinates obtained by the RTK-GPS method.

was the most accurate method (see results section).

cover the total study area and allow the evaluation of the DSM.

The results of the comparisons led to the selection of the point cloud-based methods and the rejection of the photogrammetric methods, since they were insufficiently dense to generate DSMs that could be guaranteed to detect possible collapses. However, the photogrammetric backpack MMS

The data obtained with the backpack MMS define the tracks from which to obtain the point cloud covering the terrain. These tracks were used as a common element to establish the zones that

In a second phase, the results from the two massive data recording techniques are statistically analysed to compare the numerical and graphic products obtained and to define the DSM more clearly. As has been mentioned, a supported network of control points was defined with RTK-GPS to serve as the basis for the UAV flight and the MMS backpack. This network established the real precision of both models and allowed the study of parameters such as the distance to the track in the case of backpack MMS, the point density according to the method used, and the veracity of the model with regard to walls, roofs, steeply sloping areas, and other elements. These points acted as a geometric control of the parameters to be assessed. The methodology applied in each system is shown

**Figure 5.** Workflow of the methodology for data acquisition where (**1**) refers to data acquisition and post-processing for backpack mobile mapping, UAV, and RTK-GPS, and (**2**) represents DSM processing and stability evaluation. Note the interactions between the benchmark survey (GPS) and backpack mobile mapping and UAV surveys. **Figure 5.** Workflow of the methodology for data acquisition where (**1**) refers to data acquisition and post-processing for backpack mobile mapping, UAV, and RTK-GPS, and (**2**) represents DSM processing and stability evaluation. Note the interactions between the benchmark survey (GPS) and backpack mobile mapping and UAV surveys.
