[CHAP.1] Reality Capture in construction: "As-Built" 3D Modeling

Chapter 1

 

"Reality Capture" encompasses a series of technologies that capture the reality of a construction site using LiDAR and/or images. By combining this data with other project design information layers on the same platform, a solid geospatial information base is established, facilitating understanding and analysis in construction projects.

Reality Capture focuses not only on data acquisition but also on its processing and how all this geospatial information is shared in collaborative environments, with the aim of creating "As-Built" models, controlling the geometric quality of elements materialized on-site, optimizing planning tasks, and monitoring project progress, among others.

 

Figure 1. "As-Built" 3D model and point cloud.
Reference: Geocom and Bim Consulting

 

In this experience, we delve into the Trimble Reality Capture workflow to generate an "As-Built" 3D model of a 5-story building with 2 underground levels.

From initial capture to the creation of the final 3D model, we explore how these digital tools enable an accurate and detailed representation of existing conditions, offering a robust solution for construction project documentation and management.

The experience is divided into 7 chapters where we will review the Trimble Reality Capture workflow:

1. Point cloud capture and registration - Trimble Series X / Trimble Perspective
2. Conventional topography for cloud referencing - Part 1: GNSS SIRGAS linkage, UTM Projection
3. Conventional topography for cloud referencing - Part 2: Traverse with total station
4. Point cloud processing - Management, filtering, classification, orthophotos, among others.
5. "As-Built" 3D Modeling – Trimble Realworks/Trimble SketchUp
6. Trimble Connect - Common data environment (plans, design 3D models, "As-Built" 3D model)
7. Trimble Sitevision - Augmented reality system for design and inspection

 

Chapter 1: Point cloud capture and registration - Trimble Perspective/Trimble Series X

In projects involving infrastructure maintenance, renovation, or expansion, a lack of geometric information is often encountered. The absence of "As-Built" plans and even design plans is common, especially in facilities that have been in operation for several years. In cases where this information exists, it is usually incomplete, outdated, and does not reflect reality with the necessary precision. Therefore, it is essential to capture reliable and accurate data of the current condition to generate the necessary documentation that will serve as the basis for new projects.

This chapter addresses the first two steps of the Trimble Reality Capture workflow: capture and registration.

Figure 2. Trimble Reality Capture workflow.
Reference: Geocom

 

• Capture Planning

The building covers an area of 100 x 30 meters and consists of 7 levels. An estimate of the necessary time on-site is made, considering relatively clear floor levels and scanner positions every 10 meters. Therefore, it is projected that around 300 scanner positions will be required. The time for each scanning position is defined based on the resolution needed for the project. For this experience, the primary interest is the condition of the rough work (slabs, walls, columns, and beams), so a high level of detail for representation is not required.

 

Figure 3. Building to be surveyed by laser scanner.
Reference: Google Earth Pro

Trimble X9, the scanner to be used in this experience, is an advanced high-speed laser scanning system that incorporates automatic calibration, auto-leveling, and an inertial sensor (IMU). The X9 provides the capability to register and reference the point cloud in the field. This increases capture productivity and reduces processing times.

 

Figure 4. Trimble X9 on a Dolly tripod for displacement on flat surfaces.
Reference: Geocom and Bim Consulting

 

For this experience, two measurement modes are used, depending on the building conditions:

• High-speed mode (1,000 kHz): Consists of 87-second scans achieving a resolution of 8 mm at 10 m. Measurement mode for distances less than 120 m.
• Indoor mode: Consists of 50-second scans achieving 15 mm at 10 m. Reduces calibration time and limits the range to 30 m for indoor applications.

The following table shows the capture times, resolution, approximate file size, and the maximum number of points that can be expected from each scan. The two measurement modes used for this project are highlighted in red.

 

Figure 5. Trimble X9 time and resolution specifications.
Reference: Trimble X9 Manual

 

Resolution is the distance between points on a component, it has to do with the detail that can be captured. Resolution depends on the measurement time and the distance to the component. The further the scanner is from the component, the lower the resolution, so a longer measurement time is required. It should also be considered that once the point cloud is registered, the resolution on a component increases with nearby scanner positions.

 

Figure 6. On the left, an 87-second scan with 27 million points, and on the right, a 50-second scan with 7 million points.
Reference: Trimble Realworks

 

 

The different resolution modes do not affect the accuracy of the point cloud, which for X9 is less than 1.5 mm at 30 m.

The project also considers the referencing of the point cloud in the field for its location in a defined coordinate system and for the "control" of the cloud itself. Trimble X9 incorporates a laser pointer which allows for the surveying of control points to reference the point cloud to a defined coordinate system. For this, it is planned to place "black and white" targets on the walls and floors, to subsequently define coordinates through a referenced total station. The targets must be positioned in such a way that they cover the scanned area planimetrically and altimetrically. The scanner positions that survey the targets must do so frontally and ideally at a distance of 10-15 meters.

Something very useful in project planning is to define "Labels". This is a tag that can be assigned to each station to allow filtering based on it later. In the project, the following labels are defined: facade, floor -2, floor -1, floor 1, floor 2, floor 3, floor 4, floor 5.

 

Figure 7. Scans grouped by tags.
Reference: Trimble Perspective

 

• Capture and Registration

The Trimble Perspective field software installed on the T10X tablet performs automatic registration in the field. This allows for on-site visualization of the point cloud, 360 panoramas, and measurements and annotations. Having the point cloud visualization in the field allows for clearly seeing areas with no data or low density, enabling complementary scanner positions. Additionally, it improves efficiency by optimizing the planning of each position, avoiding redundant positions that only make data processing heavier and slower.

 

Figure 8. Cloud registration in the field.
Reference: Trimble Perspective

  

Registration is the process of joining point clouds by means of artificial objects (spheres and targets) or natural features (characteristics of the environment itself) to generate a unified point cloud.

The registration process can be carried out in the office, through processing in software such as Trimble Realworks, Trimble Business Center, or directly in the field, thanks to Trimble Perspective field software from X7/X9. These incorporate an inertial sensor (IMU), an automatic leveling sensor, and use the same surveyed environmental characteristics to register the various scanner positions. To achieve effective automatic registration, it is important that position changes are made approximately every 15 meters. This distance depends on the characteristics of the measurement environment. The more features present, the greater the distance between positions can be, but in open and homogeneous areas, the distance between positions must be shorter. Under these conditions, it is recommended to perform more scans but with a shorter measurement time.

As spheres or targets are not being used for registration, it is essential to have enough features in the environment to meet the necessary overlap (approximately 30%) and not to tilt the equipment more than 45° when moving between positions, to increase the success of automatic registration. In case automatic registration is not successful, it is possible to use the manual registration tool. This requires the user to roughly move and/or rotate the cloud to the correct position, for the final adjustment to then be performed.

The images show the manual registration process where position 41 is the reference and position 42 is the mobile cloud. Visualization tools that only allow viewing the edges are a great help for aligning the clouds.

 

Figure 9. Manual cloud registration in the field. On the left, normal plan view, and on the right, plan view applying an edge filter.
Reference: Trimble Perspective

 

Once aligned, the "Register" button is selected, and an attempt is made to adjust the clouds. If successful, a green message appears indicating the percentage of overlap and the accuracy of the adjustment. If the message is orange, it means that the cloud adjusts but does not have enough overlap to ensure the reliability of the adjustment. Here there are two options: manually validate based on verifying that there are no double layers or a new capture to improve the overlap. If the message is red, the point cloud does not adjust, it remains misaligned. Therefore, it should be considered to capture another station between both positions to improve the overlap.

 

Figure 10. Successful manual cloud registration in the field.
Reference: Trimble Perspective

 

Regarding capture, it is planned to create a "backbone" along the project and several transversals, in order to prevent a deviation from propagating between positions that could cause double layers.

The images show the Trimble Perspective field software, where each scanner position is represented by a triangle and the green vectors represent the registrations between stations.

 

Figure 11. Cloud registration in the field.
Reference: Trimble Perspective

 

When the first scan is complete, Perspective automatically orients it along a wall it detects in the cloud, defining the X and Y axes. If you disagree, you can edit it manually with the help of a grid. As for the Z axis, it is defined by the auto-leveling sensor with less than 3" of accuracy (0.3 mm at 20m).

 

• Visualization tools

Trimble Perspective installed on the T10 Tablet shows a point cloud optimized for quick and fluid on-site viewing, which means that in this standard mode, it displays a lower point density than what was captured. However, there is a tool to focus processing power on smaller areas, allowing for the highest possible real density view.

 

Figure 12. Visualization tool to view actual point density.
Reference: Trimble Perspective

 

• Viewing box

This tool allows for the visualization of areas of interest, which can be volumes or longitudinal and transverse sections. It is also useful for visually inspecting for double layers in the registration.

 

Figure 13. Viewing box tool.
Reference: Trimble Perspective

 

• Field measurements

Having the registered point cloud allows for global project measurements. It is possible to perform vertical, horizontal, and inclined distance measurements, as well as coordinates, slopes, and perimeters. These measurements can be performed on the point cloud itself or on the 360 panoramas.

 

Figure 14. Viewing box tool.
Reference: Trimble Perspective

 

• High-sensitivity measurement

Both X7 and X9 have the ability to measure on highly reflective surfaces as well as dark surfaces. This is ideal for installations with stainless steel structures.

 

Figure 15. Measurement on reflective and dark components
Reference: Trimble Perspective

 

• Registration refinement

Upon completion of the survey, scanner positions must be refined. Previously, registration was performed only with directly linked stations. In this final refinement process, all scanner positions are adjusted relative to each other. This generates a report where it is possible to analyze the precisions and overlaps between positions, to decide if it is necessary to perform any complementary positions that strengthen the process.

The image shows the registration report provided by Trimble Perspective. A project summary is generated, indicating that there is only 1 registration set, meaning there are no unregistered stations or groups of stations. It also shows that 382 scanner positions were performed, with a registration of 0.9 mm, an average overlap of 43%, and a consistency of 97%.

 

Figure 16. Registration report.
Reference: Trimble Perspective

 

The report also provides details by position. For example, position 16 is analyzed, which has 9 links or overlaps. The report shows that with position 332, it has a precision of 5.6 mm, which must be a consequence of the low overlap between clouds, which is 21%, also causing a low consistency of 9%. Here, a new position between 16 and 332 should be considered to allow for better overlap between them.

 

Figure 17. Detailed registration report.
Reference: Trimble Perspective

 

Having a visualization of the project's point cloud in the field allows the user to ensure that all required elements of interest are captured, and the report ensures the quality of this survey.

In the next chapter of the "Reality Capture in Building - "As Built" 3D Modeling" experience, we will explore the procedure for linking a point cloud project to a global coordinate system.

 

Figure 18. Chapter 2, Conventional Topography for Cloud Referencing - Part 1: GNSS SIRGAS Linking, UTM Projection
Reference: Trimble Perspective