Point Cloud Processing: From Raw Scan Data to 3D Models

Industries such as offshore installations, refineries, shipyards, and industrial plants are increasingly using 3D laser scanning to accurately map the condition of existing facilities. However, data acquisition itself is only the first step in the process of delivering valuable results to the client.

The true value of scan data is revealed not in the field, but during the post-processing stage. Raw measurement data—usually very large point cloud datasets—must be properly processed before it can be safely used for design purposes.

The post-processing of 3D scans is a key element of digitization processes. It is at this stage that raw measurements are transformed into cleaned, correctly registered, and unambiguously positioned data in a coordinate system that engineers can rely on. Operations such as scan registration, noise removal, data optimization, and quality control ensure that the point cloud accurately represents the object’s actual geometry.

Any disruptive factors encountered during 3D scanning—such as vibrations, reflective surfaces, limited access, ongoing production operations, or moving objects—must be accounted for and properly addressed during the post-processing stage.

In industries where every minute of downtime incurs very high costs, the accuracy, quality, and completeness of the data are of critical importance. Erroneous data can lead to collisions, design revisions, or costly operational downtime. Therefore, a robust and repeatable post-processing workflow is essential. With the growing importance of digital twins and data-driven processes, well-structured post-processing of 3D scans has become one of the key elements in delivering reliable, engineering-grade data.

Definitions and Concepts

• 3D Scan – A 3D scan is a digital representation of a physical object, structure, or environment obtained by measuring its geometry in three dimensions. In industrial applications, 3D scanning is primarily performed using laser scanners or LiDAR (Light Detection and Ranging) technology, which capture millions of points with X, Y, and Z coordinates, creating a point cloud that enables the accurate mapping of real-world surfaces.
• Point Cloud – A point cloud is a digital dataset consisting of millions or billions of individual points, each with X, Y, and Z coordinates in three-dimensional space. Collectively, these points form an accurate spatial representation of an object, structure, or environment.
• 3D Laser Scanner – A laser scanner is a device that enables the precise measurement of the positions of millions of points representing the surrounding environment. The measurement is performed using a laser beam, which allows for the rapid capture of large areas by assigning X, Y, and Z coordinates to each point. The data collected by the scanner forms a point cloud.
• LiDAR - LiDAR (Light Detection and Ranging) is an advanced remote sensing technology that uses laser light to precisely measure distances to objects and surfaces. This technology enables the creation of highly accurate three-dimensional models of the physical environment, supporting the analysis and interpretation of complex spatial information.

Point Cloud Processing Workflow

• Point cloud preparation
• Registration
• Quality control
• Point cloud formats
• Applications of point clouds

Point Cloud Preparation – Noise and Artifact Reduction

In the context of laser scanning, noise and artifacts refer to unwanted points or distortions that do not reflect the actual geometry of the object or its surroundings. These phenomena are typical of raw data and can be caused by, among other things, surface reflectivity, long scanning distances, environmental conditions (dust, fog), vibrations, and moving objects.

Noise reduction is the first step in the data post-processing workflow. Its goal is to identify and remove outliers, random measurement errors, and duplicate moving objects, while preserving the actual surface geometry and sharp edges. The lack of proper filtering directly reduces the quality of the final point cloud registration and negatively impacts further applications of the data, such as 3D modeling, mesh generation, collision analysis, and dimensional inspection.

Basic filtering of raw point clouds is typically performed using software provided by 3D scanner manufacturers. Market leaders such as Leica Geosystems and Faro offer effective noise filtering mechanisms and automatic or semi-automatic removal of unwanted objects. At CAPNOR, drawing on many years of experience, we have developed our own methodology and proprietary software that further optimize this process and ensure high-quality final data.

Registration and Point Cloud Generation

Point cloud registration is the process of combining individual 3D scanner scans into a single, coherent point cloud covering the entire scope of work. Registration can be performed using control points or by aligning overlapping geometry between scans using cloud-to-cloud algorithms.

Regardless of the method chosen, a detailed analysis of both individual scan connections and the integrity of the entire registration network is essential. Local mismatches, weak connections, or low values of common geometry may not be visible on a global scale, yet they can significantly impact further applications of the data, such as dimensional inspection, modeling, collision detection, etc.

 Therefore, registration verification must include an analysis of local connections, RMS values, coverage consistency, and a visual inspection of key geometries throughout the scanned area.

Despite the continuous development of registration algorithms and the growing computational power of modern software, this process still relies heavily on the experience and judgment of the person performing the registration. Understanding how the scanner works, the placement of targets, and the limitations of the software is crucial for identifying errors that may go undetected by automated procedures. Operator supervision remains a decisive factor in assessing whether the registration result is not only mathematically correct but also geometrically reliable and suitable for further use.

Data Quality Control

Our experience shows that this is the most frequently overlooked stage in the entire data processing process, yet it is the most important. Given current production requirements and tight deadlines, it may seem that skipping this stage or performing it superficially might not be a problem. However, in our assessment, the more meticulously the quality control is conducted, the fewer potential issues with the point cloud will arise later in the project.

A structured quality control process is applied at every stage of scan processing and registration to verify the accuracy, consistency, and completeness of the data.

Quality control focuses on verifying the internal accuracy of the alignment between scans, assessed based on registration reports and RMS parameters. Incorrect alignments, excessive deviations, and poorly connected joints are detected and corrected to ensure the overall accuracy of the registration within the defined tolerance limits for the selected scanning accuracy class. These requirements may be specified by clients or by the internal guidelines of the 3D scanning company in question.

A key element of registration quality control is the verification of the overall geometry of the connections between individual scans. Each connection should have adequate overlap of the shared point cloud with adjacent scans, strong connection geometry, and, where required, reference points linked to geodetic measurements. Cloud-to-cloud connections with excessive error are excluded to prevent local deformations and cumulative fit drift across the entire connection network.

Upon completion of the internal registration, quality control is extended to include a georeferencing stage, if required. Measured control points are compared with the registered point cloud to confirm that the positioning and orientation in the global coordinate system meet the project and client requirements. Any deviations exceeding the tolerance limits result in corrective actions or partial re-registration to maintain consistency with the parent coordinate system.

In addition to geometric accuracy, data integrity and completeness are also verified during processing. This includes checking for unscanned areas, noise, unwanted artifacts, and inconsistencies in point density that could negatively impact subsequent project phases.

The final stage of quality control involves documentation and reporting. Registration statistics, accuracy classes, and a description of the actions taken are documented as part of the data package delivered to the client, ensuring transparency and confidence in the scan results. This guarantees that the delivered point cloud provides a reliable, high-quality representation of the scanned object and is ready for use in engineering processes, collision detection, and integration with a digital twin.

Point Cloud Formats and 3D Modeling Software

3D laser scanning generates large amounts of data in the form of point clouds. To ensure efficient data exchange, long-term usability, and compatibility with other engineering tools, point clouds are delivered in standardized file formats. Each format serves different purposes depending on accuracy requirements, the software environment, the scope of the project, and, above all, the client’s requirements.

One of the most commonly used formats is E57. It is an open, vendor-neutral format designed specifically for 3D data. E57 supports high-precision point coordinates, color information, intensity values, and scan metadata, making it well-suited for long-term archiving and data exchange between different software platforms. The openness of the E57 format makes it one of the most widely and readily used 3D data formats, enabling the creation of custom software solutions for editing point clouds.

The PTS, PTX, and XYZ formats are simpler, text-based (ASCII) point cloud formats. They store point coordinates—and optionally color or intensity—in a readable text structure. Although these formats are easy to read and widely supported, they result in significantly larger file sizes and offer limited support for metadata. For this reason, they are typically used for data exchange, troubleshooting, or ensuring compatibility with legacy software, rather than as final data.

In design environments based on Autodesk products, RCP/RCS formats are commonly used. These are indexed point cloud formats optimized for visualization and performance in Autodesk applications such as Revit, AutoCAD, and Navisworks. RCP files serve as project containers, referencing one or more RCS files that store the indexed scan data. Although these formats are highly effective in design and coordination processes, they are proprietary and are typically generated from raw scan formats, such as E57 or native scanner data.

Scanner manufacturers also provide native, proprietary file formats, such as FLS, ZFS, or others, depending on the hardware and software ecosystem used. This format preserves the full scanner metadata and is used during the data registration and processing stages. However, before the data is delivered to the client, it is most often converted to open or platform-agnostic formats to ensure interoperability and long-term accessibility.

In practice, the data processing workflow may rely on several formats depending on the selected 3D scanner, software, and the final delivery of data to the client or design department. Selecting the appropriate point cloud format is a key element of a data management strategy, ensuring that scan data remains accurate, usable, and compatible throughout the object’s lifecycle.

The Use of Point Clouds in Industry

The applications of point clouds are nearly limitless. The growing availability of scanning devices and ever-increasing capabilities for processing point clouds open up nearly limitless application possibilities for users. As a company operating primarily in the Oil & Gas sector, we will focus on industrial applications, while merely highlighting the potential for their use in other industries.

In the industrial sector, point clouds are used for as-built documentation, engineering support, and 3D inventory. Facilities such as oil and gas installations, refineries, power plants, and industrial plants are highly complex, making accurate documentation critical. Point clouds provide a precise geometric reference that reflects the actual condition of the facility, taking into account deviations from the design and undocumented changes.

Point cloud data supports engineering and design processes, serving as a reference for 3D modeling, modernization planning, and verification of plant layouts. Engineers use point clouds directly or convert them into 3D models for collision analysis, gas flow analysis, safety assessments, equipment replacement, and as-built surveys. In offshore environments and at brownfield sites, this reduces the need for site visits, lowers risks to personnel, and improves design accuracy.

Another important application is dimensional control. Point clouds enable precise measurements, alignment checks, fit verification, and tolerance analysis of structures, pipelines, and equipment. This is particularly important in prefabrication, module assembly, and life-extension projects, where dimensional accuracy directly impacts costs and project schedules.

Point clouds also play a key role in digital twin platforms and data management systems, where verified point cloud datasets are integrated with documentation, engineering models, and operational data. This enables the creation of a digital representation of a facility that supports maintenance planning, inspections, and long-term data reuse.

In construction and infrastructure projects, point clouds are used to monitor work progress, perform quality control, and validate as-built conditions. Scan data is compared with design models to assess construction accuracy and detect deviations early on. Infrastructure facilities, such as bridges, tunnels, roads, and rail lines, particularly benefit from point cloud-based documentation due to the scale and complexity of their geometry.

In architecture and the built environment, point clouds support building documentation, renovation projects, and the preservation of cultural heritage. Accurate scan data enable architects and designers to work with reliable geometry, especially when archival documentation is incomplete or outdated. In conservation applications, point clouds provide a non-invasive method for documenting historic structures while maintaining a high level of geometric detail.

In mining and surveying, point clouds are used for volume calculations, deformation monitoring, and terrain analysis. Repeated scans enable the analysis of changes over time, supporting safety assessments and operational planning. Environmental applications include, among others, shoreline monitoring, landslide analysis, and forest surveys, where point clouds provide detailed spatial data for the natural environment.

Point cloud technology is also used in transportation, the automotive industry, robotics, and virtual reality, where accurate spatial data is crucial for simulation, navigation, and visualization. In urban planning and smart city initiatives, point clouds serve as the geometric foundation for large-scale city models and spatial analyses.

Point Cloud Services in Capnor

Capnor specializes in acquiring and processing 3D laser scan data, particularly large-scale projects involving thousands of scans taken at a single site, which naturally presents significant challenges during the data processing and registration stages.

In recent years, rapid advancements in hardware and software have significantly expanded data processing capabilities. While this progress has improved workflow efficiency, it can also create the misleading impression that data processing is a simple and fully automated process. In practice, obtaining reliable, high-quality results still requires experience, proven procedures, and expert supervision.

As a company with over 20 years of experience in the field of scan acquisition and processing, we have developed internal procedures and software solutions that optimize data processing while maintaining the quality of the final data. The involvement of our employees at every stage of the project lifecycle—from planning and field scanning, through data processing and quality control, to the delivery of final reports, registered point clouds, or 3D models—enables us to continuously grow and reaffirm our status in this rapidly evolving industry.