How to Process Point Clouds? A Must-Read for Beginners — 5 Practical Tips Useful On Site

What are point clouds? They look difficult, but they’re actually simple!

Lately, on construction and surveying sites you’ll increasingly hear the term point cloud data. Point cloud data is 3D data that records real-world objects or terrain as a collection of countless points. While a photograph represents 2D information as a collection of pixels, point cloud data digitally represents object shapes as a collection of points with three-dimensional coordinates. What’s remarkable about point cloud data is that it can record the site’s shape as-is, faithfully reproducing complex shapes that traditional drawings cannot fully express. Even without drawings, you can create a 3D model from the real object and later cut sections or take measurements as needed, making it highly flexible. Because point cloud technology gives you a digital copy that closely resembles the real thing, it is attracting growing attention in construction and civil engineering.

It may seem difficult at first glance, but if you understand the procedure, it’s a technology even beginners can handle. In particular, recent advances allow easy 3D scanning with familiar tools like smartphones and drones, making point cloud processing much more accessible. With the momentum from the Ministry of Land, Infrastructure, Transport and Tourism’s *i-Construction*, digital surveying on site is likely to spread even further.

For example, using the LiDAR sensor built into the latest iPhones, you can simply point your phone to record the space in front of you as a point cloud. By processing the acquired point cloud data with dedicated apps or software, you can create realistic 3D models that reproduce the site’s conditions. Although “3D scanning” and “point cloud processing” may sound technical and daunting, we’re in an era where anyone can easily acquire and use 3D data if they understand the basic workflow.

This article explains the basic flow of point cloud processing for beginners and introduces five specific tips that are useful on site. I’ll try to simplify technical terms as much as possible, so if you’re thinking “I want to try point clouds!”, please refer to this article. Now, let’s go through the basic steps of point cloud processing in order.

Workflow of point cloud processing: from on-site scanning to data acquisition, processing, and utilization

The general workflow of point cloud processing can be divided into three steps: (1) on-site data measurement (scanning), (2) processing and editing the acquired data, and (3) utilizing the data. Let’s briefly look at each step below.

1. On-site scanning (data acquisition): First, scan the target objects or terrain on site to acquire point cloud data. Traditionally, measurements were typically made using high-precision tripod-mounted laser scanners (TLS) or drones for photogrammetry, but today you can easily acquire point clouds with mobile scanning using a smartphone. For example, with LiDAR-equipped iPhone or iPad models, you can launch a dedicated app and slowly move the device to 3D-scan the interior of a building or areas around structures in a matter of minutes. For measuring large outdoor areas, you can also obtain point clouds from the air by mounting lasers or cameras on drones. The important thing is to capture the target without omission. Scan with appropriate procedures according to the equipment and method, and be sure to acquire point cloud data for the area you need.

2. Data processing and editing: Next, process the acquired point cloud data using software on a PC or cloud service. If you scanned from multiple positions, perform alignment (merging) to integrate the point clouds into a single coordinate system (this step is called “registration” with laser scanners). Also perform pre-processing such as removing unnecessary noise points and thinning out the data (increasing point spacing) if the data volume is too large. By importing point clouds into dedicated point cloud processing software or CAD, you can perform various analyses such as creating cross-sections or overlaying with design data to detect discrepancies. Recently, some cloud services automatically process and analyze point clouds just by uploading them, so you can handle point cloud data easily even without installing expensive software on your PC. Note that point cloud data can be very large, so a computer’s specs—such as memory and GPU—are important for high-precision analysis. If your PC becomes sluggish during processing, one option is to rely on cloud services instead of pushing your local machine.

3. Data utilization and sharing: Once processed, point cloud data can be used for many purposes as site records or surveying deliverables. For example, you can measure distances and areas from acquired point clouds, or compare point clouds before and after construction to calculate excavation and embankment volumes. On construction sites, it’s increasingly common to overlay design models on point cloud data to check as-built conditions and perform error checks. Another advantage is that point cloud data is easy to share as a 3D model. Even without specialized software, you can inspect data using free point cloud viewers or display point cloud models on a smartphone or tablet and show them overlaid on the real scene using AR (augmented reality). If you store data on cloud storage on the internet, you can instantly share freshly acquired point clouds with colleagues in the office by sending a URL, enabling remote joint inspections. Point cloud data can also be stored as a site digital archive and used for later maintenance or renovation planning. In practice, you can refer to point clouds after project completion to observe long-term changes or to understand existing structures during remodeling, making them valuable long-term data assets. By understanding the entire workflow from measurement to processing to utilization, you should be able to grasp the overall picture of point cloud processing.

Five tips for beginners that are useful on site

From here, I’ll introduce five concrete tips beginners should keep in mind when performing on-site point cloud measurement and processing. Small tricks can greatly improve data quality and work efficiency, so please try them.

• Tip 1: How to walk and move your smartphone during scanning (stability and overlap) For point cloud scanning, stable movement and ensuring overlap in the captured area are important. When measuring while walking with a smartphone or handheld scanner, move smoothly at about half your usual speed. Sudden changes in direction or frequent turning can cause data instability, so gradually change the camera (sensor) orientation. Hold the smartphone firmly with both hands and stabilize your arms to prevent blur. Also, instead of trying to scan a vast area all at once, divide the area into sections and cover the whole area by walking in a zigzag pattern to reduce missed spots. If you leave and later return to the same place after a long time, the system may recognize it as a different location, causing points to be recorded twice (ghosting), so be careful. If you notice missing areas, immediately go back to fill them in to reduce duplication. When measuring floors or ground surfaces, point your smartphone about 30–45 degrees downward relative to the ground to avoid missing points. When scanning vertical surfaces such as walls or cliffs, face the target as directly as possible so the laser reflects properly and the point cloud is densely captured. The basic rule is to scan with “stable posture, slow movement, and routes planned for overlap.”

• Tip 2: Watch brightness and reflections! Shooting conditions for outdoors and indoors Pay attention to the environmental conditions when acquiring point clouds. LiDAR-equipped sensors work in low light, but they are not good with excessive direct sunlight. In outdoor conditions with strong sunlight, such as at noon in midsummer, the sensor may pick up infrared noise and measurement accuracy can degrade. If possible, scan on cloudy days or during morning/evening hours when sunlight is softer for better results. If you must scan under strong sunlight, try shading the smartphone with an umbrella or board, or adjust the LiDAR sensor’s exposure compensation setting. Also be aware that extremely reflective or transparent objects are difficult to capture in point clouds. Mirrors, glass, and water surfaces may not return lasers properly, resulting in few points or noisy, incorrectly positioned points. Very dark objects—such as black walls or people wearing black clothing—can also absorb light and cause missing points. Accept that such parts may not appear in the point cloud and record them by other means (e.g., conventional photography) when necessary. When capturing color point clouds indoors, ensure adequate lighting because the camera cannot record colors in complete darkness.

• Tip 3: Upload and organize data to the cloud immediately After scanning, quickly save and back up the data. Point cloud files are large (one measurement can result in millions to hundreds of millions of points, with file sizes from hundreds of MB to several GB), so losing them due to smartphone or PC failure can be disastrous. For internal sharing, cloud storage is more convenient than handing over USB drives. For example, if you upload data from your smartphone to a company cloud folder immediately after acquisition, you can simply share a URL for office members to view and download the data without being on site. As data accumulates, management can become complicated, so it’s important to separate folders by project and include dates and locations in file names to stay organized. This prevents situations where you later wonder “Which site is this point cloud from…?” Uploading point clouds to the cloud can also automatically generate 3D viewers so you can intuitively inspect data in a browser. Make good use of the cloud to prevent data loss and facilitate smooth information sharing.

• Tip 4: Check on site immediately after acquisition (prevent omissions and misalignments) If you realize “I forgot to capture this!” after leaving the site, it’s usually too late. Make a habit of checking point clouds on site immediately after measurement. Thoroughly inspect the generated 3D model on your smartphone or tablet right after scanning to check for missing areas or obvious misalignments. Small details can be overlooked on a phone screen, so if possible view and rotate the point cloud on a larger device like an iPad. If there are gaps with no points at all or parts where the shape is clearly distorted (for example, where the system lost its pose during scanning), perform additional scans or retakes on site. Discovering problems only after returning to the office is too late. Especially check that key dimensions and important structures are properly measured by using distance measurement features on site to verify reference dimensions. Many modern apps complete data processing on site and allow immediate checking of point clouds, so make use of them. Remember that “capture” is not complete until you “capture and verify on site.”

• Tip 5: Use AR display to intuitively check differences with the design! Make the most of acquired point cloud data by using AR features on site. With a smartphone or tablet’s 3D display capability, you can overlay the scanned point cloud or a 3D model created from it onto the real scene. For example, display the point cloud model of a captured structure together with the 3D design model in AR at the site. The design model will appear “floating” over the real scene on the screen, allowing an intuitive comparison between the as-built condition and the design. Overlaying models on site makes it instantly clear “where the errors are” and “how far construction has progressed,” more so than comparing drawings or point clouds alone. AR visualization is also an excellent tool for intuitive on-site communication. In construction management, it’s increasingly common to hold up a tablet to overlay the expected completed view on the actual site, making it easy for anyone to visually understand. The operation is simple—press the “AR display” button in a compatible app—so try incorporating AR into your on-site checks.

Summary: With smartphone-mounted LRTK, anyone can easily perform point cloud surveying, AR checks, and cloud reporting

Finally, I want to emphasize again that we are entering an era in which anyone can easily perform point cloud surveying. Until now, 3D point cloud measurement required specialists and advanced equipment, but today a smartphone and a little know-how allow site staff to carry out 3D surveying themselves. As introduced in this article, by paying attention to scanning posture and environment and by sharing and verifying data immediately, even beginners can acquire high-quality point cloud data.

In recent years, solutions have appeared that dramatically improve positioning accuracy using RTK-GNSS receivers attachable to smartphones. For example, by using a smartphone-mounted RTK rover device called LRTK, you can add centimeter-level positioning information (half-inch-level positioning information) to point clouds acquired with a smartphone, enabling high-precision surveying that used to require specialized equipment to become accessible to anyone. With a smartphone + LRTK setup, you can complete the entire workflow on site—acquiring 3D point clouds, performing as-built AR checks, and reporting via cloud sharing—all one-stop on the spot. For example, even on sites with complex structures, one person can walk around with a smartphone to complete detailed point cloud surveying, immediately overlay the design model to check construction accuracy, and quickly share results to the cloud to report to supervisors or clients. Thanks to these accessible technologies, the once-daunting task of point cloud processing has become far less intimidating. Take this opportunity to use the latest tools and try 3D point cloud measurement with just a smartphone—it will surely improve site work efficiency and accuracy dramatically. Make point cloud processing your ally and start making your future site work smarter!