Use of Advanced Technologies for Topographic Surveys in Civil Construction

,


Introduction
The complexity that has characterized market relations in recent years, with the demand for product and process innovation, has even had repercussions in more traditional activities, such as in the civil construction segment (Cavalcanti et al., 2018;Alaloul et al., 2019). Despite being traditional, this segment has undergone several changes in the business model, as a result of important technological advances, requiring companies to adapt to new conditions by incorporating new technologies, professional qualification, materials and developing new processes, considered innovative, resulting in projects increasingly complex (FIRJAN, 2014).
3D modeling uses software to create a mathematical representation of a three-dimensional shape. According to Alves (2018), some modeling software allows 3D models to be shared and viewed anywhere, especially at the construction site. In this way, the project can be changed or updated in real time, inaccurate data and calculations can be hastily corrected and the company avoids rework in several steps that would end up generating extra costs and delays in execution.
In this technology environment, we cannot forget BIM. According to Wilson and Heng (2021), BIM is changing traditional construction practices in a broader sense, in terms of people, processes, work, culture, communication and business models. According to Gu and London (2020), BIM involves the application and maintenance of an integrated digital model of all construction information at different stages of the development life cycle in the form of a data repository, including geometric and non-geometric information.
The ability to share information and experience, obtain true cost estimates from the outset, identify problems and implement solutions based on reliable information prior to construction all benefit from saving time, money and achieving a superior result.
Regarding the challenges in the Brazilian context for the use of these technologies, the country faces difficulties such as lack of investment in equipment to incorporate these technologies, adaptation of layouts, processes and forms of relationship between companies in the production chain, as well as the difficulty of creating of new specialties and the development of new technologies (CNI, 2016). In addition, there is a need to adopt relatively quick measures to avoid a competitiveness gap between Brazil and some countries where Industry 4.0 has already started to become a reality.
To obtain the data that will support these technologies, we highlight the three-dimensional survey using Terrestrial Laser Scanner (TLS), which generates a cloud of dense and high-precision points. This product is widely used for taking dimensions in the three orthogonal cartesian axes. Because it uses laser measuring equipment, applying surface scanning, it has a high economic value. In this way, technical and scientific research seeks simplified methodologies for generating models, equivalent to those generated by the TLS (Inocencio et al., 2014;Young & Seonghyuk, 2019).
As alternatives to the TLS, there is the Unmanned Aerial Vehicle (UAV), popularly known as a drone, which through imaging and specific software is capable of generating three-dimensional models equivalent to the models generated by the TLS. However, Fonseca e Silva and Maia Gomes (2020) in their results points to the need for further research with regard to positional precision and accuracy, the feasibility of applying UAV imaging in civil structures, as well as producing equivalent end products of this technology (digital terrain model and point cloud).
To solve issues like to reduce the digitization time and costs while maintaining a high coherence between the physical artefact and its digital counterpart, in a context where the purpose is to massively digitize complex objects in their original setting by minimizing the impact, several methodologies are being discussed to improve the efficiency of digitization regarding image capturing, 3D model creation, scaling and mesh editing. This study is corroborated by Medeiros, Figueira and Vasconcelos (2022), where they highlight that, despite the merit of the scientific community turning to studies on AR in recent years, joint efforts are needed to solve the challenges faced by technology, such as the parallax, hardware limitations, the high consumption of time to configure the markers, data storage limitations, poor georeferencing and inaccurate tracking of the environment, in order to make the use of AR even more assertive, efficient and accessible in the construction site.
In this context, the research question is how UAV can replace the use of TLS, obtaining accurate results. This article aims to evaluate the feasibility of replacing the use of a three-dimensional TLS by the use of UAV in three-dimensional surveys for civil construction. For this, it presents a comparison in the use of equipment for carrying out three-dimensional surveys in civil construction, using as an example the UAV and the TLS, in addition to the total station-more conventional equipment-for surveying the control points.
With the results obtained in this comparison, it is possible to manage the positive and negative points of each equipment, where each one is more viable depending on the case to be studied. In its development, research was carried out on the functioning of the equipment mentioned above, as well as the precision of each one and the recommendations contained in the norms regarding the acceptable standard deviations. After collecting this information, for the methodology, a field study was carried out in a real work still in progress within the municipality of Gravatá-PE, where it was possible to use the total station together with the Global Navigation Satellite System (GPS/GNSS) RTK-auxiliary equipment, TLS and VANT to carry out the same planimetric survey, and with the use of software, results were obtained that served as comparison instruments.

Perspectives of Imaging Technologies in Brazil
The use of TLS has expanded a lot in recent years in the field of graphical and metric documentation of objects, mainly because it is a non-destructive and non-invasive technique, which does not involve direct contact. TLS is Remote Sensing equipment that make it possible to collect a large number of data from the observed surface, with high precision and a fast acquisition rate (thousands and even millions of points per second (Inocencio, 2014;Kerle, 2019).
The basic operating principle of the TLS is to measure the time required for a laser pulse to travel from the transmitter to the reflective surface of the target and back to the receiver. The light beam emitted by the equipment travels through the atmosphere and interacts with the target object. The constituent atoms and molecules of the target reflect or absorb electromagnetic radiation and its backscattering gives rise to remote laser detection (Becker, 2019). The files generated by TLS is based on a structure where the coordinates of the points in space (x, y, z) are stored, the laser pulse return intensity value (I) and, if available, the values from the digital camera attached to the equipment. The final product of a scan is a cloud of points with spatial coordinates and their corresponding intensities, forming a 3D image of the scanned structure (Ferraz, Souza, & Reis, 2016;Becker, 2019).
This technology is not recent in civil construction and has been used, among other applications, to estimate the deformation of arches and vaults, based on the symmetry of cuts obtained along the vault guideline Cintra, 2017) , obtaining as-built designs (Bosché, 2010;Klein, 2012;Miranda, 2020;Gouveia, 2021), automatic recognition of surface damage related to mass loss (Teza, Galgaro, & Moro, 2009), scanning for reconstruction models of building facades and pathological manifestations in materials that make up the construction of buildings (Pu & Vosselman, 2009;Ballesteros, 2020;Ballesteros & Lordsleem Junior, 2021), efflorescence in granitic rocks on the walls of buildings, through the images generated in the scan (Armesto-González et al., 2010) and even to detect the proliferation of mosses in reinforced concrete structures (González-Jorge et al., 2012).
To use the UAV as a solution, it is necessary to apply the principle of photogrammetry and aerial surveying. Objectively, photogrammetry is the science or art of obtaining reliable measurements through photographs. In order to analyze the data obtained in photogrammetry, it is necessary to know the phototriangulation technique, which is presented as a technique that helps in the mathematical interpretation of photographs. The aerial survey, on the other hand, can be described as a set of air or space operations for measuring, computing and recording terrain data using specific sensors, consisting of an aerospace phase for capturing and recording data and a phase that refers to the data processing (Macedo et al., 2020;Sobrinho, 2021).
In order to measure and confirm the results, it is necessary to implement control points on the ground, these points are targets or georeferenced objects on the ground that will appear in the aerial images, that is, photo-identifiable. These control points are used to correlate the image coordinate system with the terrain coordinate system. They are reference points on the ground that are used in the post-processing of the images to increase the accuracy of the final products generated (Neto, 2015;Cintra, 2017;Malik & Guidi, 2018

Materia
To achieve the first ste field, for d carrying ou In the third the last ph developed y with a total s ce, from E1 th uded, the total J ased on the ava ated in the city apital of Pernam itants (IBGE, 2 olidays, many i eople in the ci ion between t can be found 005), influence re abundant ra the coldest mo  the points on the boundary of the terrain and the control points on the front of the terrain were irradiated. The irradiations were made using prism with direct reading in the work.
After obtaining the georeferenced coordinates of E1 and E0, the data were processed in the Autocad Civil 3D software, where the main traverse and all the irradiations were calculated, and through this process the precise coordinates of all the vertices of the work were obtained, including the control points.
The equipment used as a UAV was a DJI Mini 3 Pro (DJI RC) (GL), equipped with a 20-megapixel camera and a precision GPS/GNSS kit (PPK-Post-Processed Kinematic) Emlid M+ (L1) with its respective Emlid RS base (L1). Although it is not the most used model in the literature, it was chosen because it is a low-cost device.
The overflight with the UAV was carried out with the aid of the Matterport MAP Pilot application, the mission was of the Oblique type at 25 m in height, with the camera directed at 60° in relation to the nadir. The overlapping of the photos was 89 and 90% (overlap, sidelap).
As TLS, the Matterport Pro2 was used, which has a 3D sensor of structured light (infrared), 20 seconds of capture time per scan, 99% accuracy within range and a maximum range of 4.5 m. Depth resolution is 10 points per degree (3600 points at the equator, 1800 points at the meridian, about 4 million points per panorama).

Step 3-Accuracy Analysis
The data collected in the field were treated and analyzed in the laboratory, using software such as Guandalini PPK, AutoCad Civil 3D, Agisoft Photoscan, in addition to Map by Matterport. In this step, the topographic data of the control points collected in the field were treated in the AutoCad Civil 3D software and their coordinates corrected. The precise coordinates of the photos, which were obtained through the GPS/GNSS onboard the UAV (Emlid M+), were treated and corrected through the Guandalini PPK software, while the images collected in the field were treated in the Agisoft Photoscan software, in which, through algorithms, applies to the phototriangulation technique: aligning, georeferencing and generating the specific products of aerial photogrammetry.
The photographs obtained in the aerial survey (445) were processed with reference to the 10 control points surveyed in the field, and the final product was an orthophoto in '.tiff' format, georeferenced, which was imported into the AutoCAD civil 3D software.
The use of technological equipment, such as the total station, for measuring measurements within civil construction requires some care so that the results obtained do not suffer variations that do not meet the pre-established standard deviations in NBR 13133 (Matos, 2018). What differs them is the way they are handled, and with the laser scanner and the total station, many processes are done manually, for example, the correct use of the prism in the total station, which needs to be positioned so that the emitted laser through the device directly hit your target. In addition, it needs to be level with the ground, which is why it usually has a spirit level in its structure, which makes the operator precise when holding it. With drones, most processes are programmed through a device, which can be the user's own smartphone, and thus the drone does the service automatically according to the programming made, which can leave margins for error if the programming is not done correctly. correctly (Sobrinho, 2018). Matos (2018) shows that the technical unpreparedness to use this equipment can directly interfere with the final project, because with the mistakes made, it may be necessary to redo the fieldwork. In addition to the technical preparation for their use, it is also necessary that the conditions of the device are in accordance with the compliances so that the work is carried out with quality.
According to Rebelo (2019), no matter how hard you try to reduce and even eliminate errors, they always happen. The most common errors occur on three occasions, errors due to natural factors such as wind, humidity, fog, etc.; errors due to instrumental factors, which refer to defects in the device used; and errors due to personal factors, which, as previously mentioned, occur when the operator does not perform the work accurately. The latter may still be related to the user's sensory issue, such as touch and vision, each of which has its own annuity.
Still within the scope of possible errors, Matos (2018) showed in his work that with the use of the total station the influence of the ambient temperature interferes with the results obtained, and the heat of the sun acting directly on the device caused the measurement to suffer a error of approximately 3cm, not being within the standard established by the norm considering the type of device used (FOIF station model RTS/OTS 685). As previously mentioned, NBR 13133 provides a table regarding acceptable standard deviations in the use of total stations, according to the class of each equipment.
the results obtained with UAV also depend on a number of factors to be as accurate as possible. The lack of jms.ccsenet.org Journal of Management and Sustainability Vol. 13, No. 2; knowledge in its handling and the interference of external factors are examples of these factors. According to Xavier (2020), for the results to be more accurate, it is necessary to combine some equipment together with the drone. The RTK GPS is the equipment that helps the drone to carry out the surveys, obtaining the coordinates (X, Y and Z) of the points of interest. Also, according to Xavier (2019), the GPS/GNSS system integrated into the drone does not have good accuracy for aerial surveys, and can generate errors within a radius of 5 to 10 meters from the point of interest depending on the type of device used. Arias (2017) points out that another factor that directly interferes with the results obtained with drones is the solar trajectory, because depending on the time of day, the shadow area can make it difficult and generate errors in image processing.

Results
The search results are presented below.

Survey and Analysis of Control Points
The control points, to verify the accuracy, were surveyed using the Total Station and are shown in Table 1, together with the RMSE resulting from the processing. At the end of processing, a report was generated with the Root Mean Square Error (RMSE) values of the points. This RMSE is the root mean squared error of the difference between the prediction and the actual value and explicitly represents what various methods tend to minimize (Lopes & Barbosa, 2020). The result of adjusting the three-dimensional model, using the control points, can be seen in Table 2. The processing resulted in a three-dimensional model, point cloud, with 7,490,386 points, that is, a density of 1,706,238 points/m². Figure 4 shows the three-dimensional model obtained through imaging.

jms.ccsenet
The proce showed an componen field. This the field, a

Accura
When deal cm) for the that the re noteworthy an importa Another co both in the 3D. Comp

Discussion
It was evident in this study that the ambient lighting is an important issue to be evaluated in the planning and execution of the imaging. The shaded locations in the model showed greater deficiency of details for both equipment.
For aero photogrammetry, flight altitude is an item of great relevance. In this way, it could not be different for the final result of UAV imaging. The richness of detail in the final model directly depends on the amount of detail observable in the photos, that is, it is necessary to correctly equate the camera resolution ratio with the flight height and the desired richness of detail.
Another issue regarding the use of the UAV in this type of application is that it is restricted to use in uncovered environments, preventing its use in closed and covered environments, due to the need for visibility for the GPS/GNSS, both for navigation of the drone and the on-board precision one (PPK Emlid), while the LST model adopted allows imaging in closed environments with low light.
Equating the point cloud generated by UAV imaging with a probable modeling using TLS in an equivalent area, the number of points generated is greater, but even so it is reasonable to inform that the number of points to be generated in modeling by imaging can be adapted, being defined by the user while executing the processing in Agisoft Photoscan. It is noteworthy that the increase in the number of points in the processing phase is due to interpolation processes and does not represent new points obtained in the mapped object, which can compromise the accuracy of the survey.
Still on a direct comparison between the methods, regarding the time taken in the field information, the time in the imaging method was 58 minutes, from the implantation of the points until the end of the overflight. In the LST scan, the time was 3 hours and 18 minutes. This shows that the imaging method presents greater simplicity in its field conception.
In terms of processing, imaging made great demands on the machine (computer) used. Even though it is a specific equipment for processes that demand high processing capacity (Central Process Unit-CPU) and memory (Random Access Memory-RAM), this processing required 17 hours in the Depth Map and 9 hours and 27 minutes in the Dense Point Cloud, remembering that such processes were continuous, without turning off the machine during this entire period. Still comparing with the LST, the processing of the points obtained with the laser scanner demanded 5 hours and 30 minutes.
It was characterized that in the imaging method the final result is intrinsically linked to the ability of the software algorithms to resolve possible dimensional errors in the model, given that the algorithm interprets repeated overlapping of images of the same scenario, seeking homologous points. In LST scanning, however, this resolution is directly linked to the robustness and precision of the laser equipment and not to the processing software.
In order to carry out modeling as a basis for projects in civil construction, it is clear that it is necessary to apply imaging techniques more efficiently to obtain a more accurate product, in order to reach millimetric accuracy. Also studies on better positioning of targets and georeferencing of models would be of great value.
As observed in the referenced studies, the TLS is able to detect moisture, biodeterioration and cracks, pathological manifestations, in addition to allowing the visualization of dimensional changes and deformations in the order of a few centimeters. By enabling the visualization and scanning of a structure without direct contact with it, with a range of meters and even kilometers, the LST would be an option for inspections in civil construction. Image processing techniques and classification algorithms can also be used to create a pattern for the incidence/intensity/amount of damage to structures, which in current methodology depends on the inspector's qualitative criteria.
This study is in its initial phase and opens up a range of opportunities in terms of scientific development and innovation, bringing together the Remote Sensing area and Civil Construction. It is believed that with the advancement of the technological capacity of the equipment used here together with their respective software, 3D modeling through these new tools, replacing an LST, will soon have a viable and reliable methodology to apply in three-dimensional modeling.