Point clouds are the raw data of 3D Laser Scanners. It is the data on which measurement evaluation is made and also deliverable products are developed. Therefore, processing the obtained point clouds with various tools is as important as obtaining point clouds by making measurements in laser scanning projects. While these properties are the parameters that directly indicate their quality and ease of processing, they all have some general characteristics of the point clouds. (Aygun, 2021)
Practitioners have many possibilities when working with point clouds thanks to the properties of the data:
- Point clouds can be rotated as desired on the software and can be seen from different perspectives and distances,
- Measured points in the cloud that are not needed or desired can be easily deleted,
- Control measurements can be made easily by measuring distances in the point cloud data,
- CAD objects can be created from relevant parts or all of the point cloud, thanks to surfaces that match the real object very well. This feature represents the most important feature of the point cloud; the door to reverse engineering for all structures has been opened with this data feature. As a result, the export of laser scan data to a CAD system is made possible by the modeling process. In this way, an economical way has been found that can export building geometry or industrial plant inventories to a CAD system. Restoration or challenging interior planning works can be done digitally with this process now. Especially for the modernization of old buildings, the new method was quickly adopted when combined well with other laser-assisted methods (laser meter, total-station). (Aygun, 2021)
Point clouds contain two types of data:
- Metric data: Defines object geometry and shows spatial relationships between objects in the environment,
- Visual or thematic data: It is generated to describe the features of the object’s surface and can be used to calculate the reliability of the distance data for each point.
Working with 3D Laser Scanners is synonymous with working with point clouds. Examining the scanning measurements within the scope of accuracy, integrity and the information they contain determines the quality of these processes. People working with 3D Laser Scanners should have knowledge of the structure and quality of point clouds and have experience in processing point clouds.
Structure of the Point Cloud
A point cloud is a series of 3D points that define the outline or surface features of an object, and each laser scanner generates point cloud data. Along with the 3D coordinate data of the points, the intensity data of the laser beam reflected from the object surface is also recorded, and uncolored point clouds are usually presented in semi-colored form with this density data (in grayscales). Since active energy is recorded in laser scanners, intensity values can also be recorded in dark environments.
This density data representation of complex structures is suitable for visual analysis; since different surfaces have different reflectance values, objects that are difficult to distinguish with their true colors can be separated by density values. However, true color point clouds are often preferred in the final product presentation.
Figure-1: The representation of colorless point clouds is made with the recorded laser signal intensity data. Since the laser signal reflected from different object surfaces is recorded at different intensity values, it is possible to distinguish different objects through this display.
Figure-2: Different coloring approaches for a point cloud.
Point cloud data is generated in different data formats. The raw data of each laser scanner is obtained in different formats, and the majority of these formats can be converted to each other. Basic laser scanner softwares developed by manufacturers perform basic operations such as opening, examining, cleaning, and saving point clouds. Although each software is mainly focused on the point cloud format produced by its own device, they have the ability to open point cloud data with other extensions and perform the same operations.
Point clouds can be obtained from a single scan, but data from multiple scans often need to be combined (aligned) to fully identify an object. The combining process – terminologically called registration – includes the same surface data from multiple scan stations. However, some systematic measurement errors are seen together with the physical properties of the scanned object in combined (registered) point clouds. Softwares are used to eliminate these errors and to show the object surface represented by the point cloud as clearly as possible. Each software may use different algorithms.
The process of transforming a point cloud into a final product is through the sequential execution of interrelated steps. The point cloud is processed by converting it to different formats with the same geometric feature. Therefore, the raw point cloud should be as simple and inclusive as possible. This depends on the quality of the laser scanning measurement.
Quality of Point Clouds
3D Laser Scanning is a very successful measurement method in data production that meets the need for high accuracy. The absence of a standard for the quality of 3d laser scanning measurement means that there is no standard for the quality of point clouds that are the measurement product.
However, the required qualities of point clouds can be determined by the object being measured. Objects to be scanned often have complex structures or their shapes do not have a regular geometry. Point cloud representing the complex structure and presenting the desired structure information is possible by providing some technical parameters by the scanning device and scanning software.
An important step in developing a measurement strategy with laser scanners is to select the scanner that can obtain the representation of the object to be measured with sufficient accuracy. In this way, point clouds with the following characteristics can be obtained:
- Ability to show all the necessary details on the object surface,
- Having real colors to represent the object when necessary,
- Many different objects that are not required will also be measured during the measuring process. Overloading the hardware capacities of the computers used may cause incomplete or incorrect operations on point clouds. Point clouds should contain the required geometrical information of the object and its surroundings, the redundant parts should be removed from the data.
- Point clouds can be linked to geodetic control networks. In this case, if target points with known coordinates will be used, it is necessary to scan the targets with sufficient density in order to make the relevant conversion.
Processing Point Clouds
Point clouds are transformed into final products by going through various processes in software. Software developed for 3D Laser Scanning need to process large amounts of data. Point clouds that take a few minutes to measure can take hours to process.
Digital objects with defined geometry – called CAD objects – are objects to which point clouds are often transformed. The result of a single scan with a dense spacing is millions of points. Importing them directly into CAD software means overloading. For this reason, companies that develop laser scanning software develop also some special plug-ins that work in the background of their CAD software to ensure that the complete point cloud is loaded. Based on this feature, it is quite easy to obtain the desired CAD objects or perform data analysis. (Aygun, 2021)
Processing point clouds begins with analyzing individual scans. In these reviews; some of the errors caused by the scanner, user, scanned object and scanning environment can be examined. Some of these errors are fixed by basic scanning software.
Each software has special filtering tools to fix these errors but these operations may not fully meet the user’s requests. For this reason, point clouds should be checked after the procedures and the performance of the tools used should be evaluated. As practitioners try such operations with different software, they gain the ability to perform the necessary operations faster and more accurately.
Figure-3: Filtering process in Faro Scene software.
The cleaned and registered point clouds can be imported to the software where they will be prepared as the final product. The general term for the process of transforming the resulting point cloud into a more useful product is modeling, or more specifically surface or geometric modeling. There are different approaches to the modeling process.
For a small object, the most typical product would be a digital model of the object’s geometry, possibly in the form of a lattice surface such as a triangular irregular mesh (TIN – Triangular Irregular Network). (Aygun, 2021)
Registering point cloud data may require checking the model produced, depending on the complexity of the surface. For example, large flat areas require less triangles than more detailed areas. This gives the user control over the size of the produced model, preserving detail where it is needed, but removing unnecessary complexity in other areas. (Aygun, 2021)
To create a complete and continuous model of the object, it may be necessary to edit the TIN to fill in the gaps on the surfaces of the object for which no data has been collected. The resulting TIN is suitable for use in various types of analysis. More suitable models appear as improved geometric models (mesh), which are also formed by triangles. The figure below shows the model created from the point cloud obtained by laser scanning.
Figure-4: A stone historical relic (above), mesh model (below).
Registered point clouds are not delivered only as models. The regular geometric shape of the measured object allows 2D products to be presented showing object edges. Vertical and horizontal sections, plans, and facade views are examples of these data.
Figure-5: An orthophoto example created from a point cloud.
Geomatics engineers often work with professionals from other disciplines on projects that require 3D Laser Scanning measurements. In such interdisciplinary studies, it is a professional principle, and it also means avoiding time and expense, to provide the most appropriate response to the needs of the other party.
Delivering a product derived from laser scanning data may only be the beginning of the process of answering original research questions. Professionals receiving the final product will need to do some analysis using the product. Point clouds, even in their raw form, can enable some analysis to be performed, but special studies such as surface analysis require working on the resulting model.
As mentioned before, 3D Laser Scanners also record the intensity data of the laser beam reflected from the object surface. Because most scanners operate outside the visible spectrum to the human eye, the collected intensity information is often slightly different from the information seen in the field. In some cases this information can
help show the variation due to the type of material on the surface.
Figure-6: Laser scan intensity data representation of an excavated surface (left), and stratigraphic record of the same site prepared by archaeologists (right).
The vast majority of processes made on the point cloud cannot be done automatically. These processes can be quite demanding and require a personal skills. Therefore, all the operations performed within the scope of a project and the raw data received from the device should be stored as project metadata. Metadata should include information such as the name and date of the work performed, along with the technical specifications of the collected data.