Table of Contents

• What is point cloud scanning?

• Main practical challenges - Managing positional accuracy and surveying errors (absolute coordinates, relative coordinates) - Scan omissions, noise contamination, and overlooked measurement areas - Point cloud data size, PC performance, and portability issues - Software proficiency, personnel training, and operator dependency - Burden of secondary processing such as cross-sections and longitudinal/transverse comparisons - Difficulty overlaying with drawings and design data - Lack of cloud/file sharing and coordination with clients - Field conditions: rain, wind, on-site display and verifiability

• Solutions - Ensuring positional accuracy by combining GNSS or RTK - Avoiding scan omissions through shooting plans and control point placement - Lightening and routinizing data capture with mobile LiDAR and smartphone point clouds - Improving shareability via cloud processing and point cloud viewers - Combining with LRTK to enable high-precision positioning and point cloud recording with a single smartphone

• Summary

What is point cloud scanning?

Point cloud scanning is a measurement method that acquires countless points in three-dimensional space for targets such as structures or terrain and digitally represents shape by the collection (point cloud data). Each point contains X, Y, Z coordinate values (and sometimes color or return intensity), and the higher the point density, the more finely the shape can be reproduced. This technology allows complex terrain and large structures that are difficult to measure manually to be acquired in a short time, and its use is spreading in construction and surveying.

There are two main methods for acquiring point cloud data: laser scanners (LiDAR) and photogrammetry. Using a laser scanner, laser light emitted from the instrument is projected onto the target, and distance is measured accurately from the time until reflection returns or from phase changes of the light, producing millions of high-precision points. Photogrammetry, on the other hand, captures the target from various angles using a regular digital camera or a drone-mounted camera and reconstructs 3D shape by analyzing matching feature points across multiple images. Photogrammetry is low-cost and can cover wide areas while also providing RGB color information, but single-point accuracy is somewhat inferior to laser scanners. Nevertheless, these methods enable digitalization of buildings and terrain with millimeter- to cm-level accuracy (half-inch accuracy).

Point cloud scanning has recently been used widely for creating survey maps, as-built verification (post-construction shape checks), earthwork volume calculations, and infrastructure maintenance. Tasks that once took days on site can be recorded in detail in a short time using point cloud scanning, improving productivity and reducing human error. However, mastering this powerful technology in practice comes with several challenges. Below we explain the main practical issues in field use of point cloud scanning and corresponding solutions.

Main practical challenges

Managing positional accuracy and surveying errors (absolute coordinates, relative coordinates)

Point cloud data obtained by 3D scanning often have very high relative accuracy (dimensional accuracy within the point cloud) on the order of millimeters to centimeters, depending on the equipment and method. However, ensuring absolute accuracy (positional accuracy within a real geodetic coordinate system) requires care. Point clouds measured solely with laser scanners or smartphone LiDAR are recorded in a local arbitrary coordinate system, so they may be offset relative to geodetic control points or the coordinate system used in design drawings. When stitching together point clouds acquired in multiple sessions, small errors can accumulate and cause registration discrepancies. If these positional accuracy issues are left unaddressed, detailed point clouds cannot be correctly overlaid with maps or drawings, risking incorrect decisions. Therefore, on site it is essential to set known reference coordinates and implement measures to monitor surveying errors before and after point cloud acquisition.

Scan omissions, noise contamination, and overlooked measurement areas

Point cloud scanning can only acquire data within the line of sight of the laser or camera, so blind spots and consequent omissions (scan gaps) can occur. For example, the back side of a structure, recesses, or depressions in terrain may not be hit by points from a single scan location and therefore will not appear in the point cloud. Rain, dust, or strong sunlight reflections can generate noise points that scatter points where none should exist. Metal surfaces, glass, and water can reflect or transmit lasers and cause incorrect points to be generated in the surroundings. If the measurement area is not thoroughly checked before packing up, one may later find that parts intended to be measured are missing from the data. To prevent such omissions and noise, it is important to identify potential blind spots in advance and cover them by scanning from multiple viewpoints and placing additional targets as needed.

Point cloud data size, PC performance, and portability issues

High-density point clouds tend to produce very large files and put heavy loads on computer processing. Point clouds on the scale of tens of millions of points can reach several GB in size, and a standard laptop may take time to open them or may crash due to memory shortage. Especially on site, tablets and similar devices often cannot display such data smoothly, making it difficult to use the data on the spot. There is also the issue of equipment portability. High-precision terrestrial LiDAR systems can weigh several kg including tripod, and batteries add further burdens, making transport and setup time-consuming. In mountainous or narrow areas where carrying equipment is difficult, adequate scanning may be impossible. These physical constraints — “heavy data,” “underpowered PCs,” and “non-portable equipment” — form hurdles to point cloud utilization.

Software proficiency, personnel training, and operator dependency

Dedicated software for handling point cloud data (noise removal, point cloud merging, 3D modeling, etc.) is highly capable but complex to operate, so becoming proficient requires time and experience. As a result, only a few in-house staff can master it, and point cloud processing tends to become person-dependent. If knowledge is not shared with newcomers and tasks are left to specific individuals, work can stall when those people are absent, and result quality may vary by operator. In addition, the number of professionals with specialized skills in point cloud scanning and 3D software is still small, which hinders field adoption. Much of the software is foreign-made and has limited Japanese documentation, so many field engineers feel a high barrier to learning the operation.

Burden of secondary processing such as cross-sections and longitudinal/transverse comparisons

Secondary processing to extract useful information from acquired point clouds also requires effort. In civil engineering, for example, it is necessary to create longitudinal and cross-section drawings of roads and embankments for as-built inspection, or to compare point clouds before and after excavation or embankment to calculate volume differences. Extracting these from point clouds involves slicing the point cloud along arbitrary planes to trace linear features or computing differences between multiple point clouds. Extracting the required cross-sections from large point datasets has limited automation in software and often relies on manual work by staff. As a result, drawing production and quantity calculations take time, leading to situations where “we took point clouds but struggle to compile the deliverables.”

Difficulty overlaying with drawings and design data

Overlaying point cloud data with existing 2D drawings or 3D design models for comparison is not straightforward. Point clouds are collections of points and are not directly compatible with CAD or BIM object data. For example, to overlay post-construction point clouds on design CAD to check for deviations, you need an environment that can display both in the same coordinate system on one screen. Some specialized software can display point clouds and design data simultaneously and perform difference checks, but data conversion and coordinate transformation steps between software can be cumbersome and introduce errors. If client drawings or existing plans only exist on paper or in PDF, rigorous comparison to point clouds becomes even more difficult. For these reasons, field teams often struggle to link point cloud usage with traditional drawings, and voices saying “we can’t fully leverage the 3D data we measured” are common.

Lack of cloud/file sharing and coordination with clients

Because point cloud data are large and specialized, sharing among stakeholders often does not go smoothly. For example, when a contractor tries to hand over point clouds to a client or designer, files of several gigabytes cannot be sent by email, and even via cloud storage the recipient may face long download times. If the recipient lacks suitable viewing software or sufficient PC specs, the valuable 3D data remain inaccessible and go unused. Consequently, parties may default to sharing traditional 2D drawings and reports, and the detailed information unique to point clouds is not exploited. There is still insufficient infrastructure for organizational information sharing, such as real-time cloud-based data linkage or environments where clients can view point clouds in a browser.

Field conditions: rain, wind, on-site display and verifiability

Outdoor point cloud scanning faces site-specific constraints due to weather and working conditions. In rain, laser beams scatter from raindrops and noise increases, and equipment may require waterproofing, making measurement generally difficult. In strong winds, drone LiDAR may not be able to maintain stable flight, and tripod sway for ground lasers can reduce accuracy. Strong direct sunlight can cause contrast irregularities in photogrammetry images, degrading feature-point matching. Additionally, confirming the acquired point cloud immediately on site is challenging. Some laser scanners provide preview via built-in screens or tablet linkage, but these are usually insufficient for full verification, and often errors or omissions are not discovered until post-processing at the office. If issues are unnoticed on site and only discovered later, re-measurement becomes necessary and is inefficient. Without adequate measures for weather/environmental factors and immediate verification workflows, stable operation of point cloud scanning is hampered.

Solutions

Ensuring positional accuracy by combining GNSS or RTK

The most reliable way to improve absolute coordinate accuracy of point cloud scans is to combine them with GNSS positioning (e.g., GPS) or RTK to provide reference coordinates. For terrestrial laser scanners, placing known control points (survey benchmarks) on site before measurement and putting reflective targets or calibration markers nearby allows the acquired point cloud to be aligned to those reference points in post-processing. For drone aerial photography or mobile scanners, equipping the device itself with RTK-GNSS enables real-time position correction during flight or movement while recording point cloud data. For example, an RTK-equipped drone can assign centimeter-level latitude, longitude, and altitude to each photo or point while flying over several hectares. Traditionally multiple ground control points (GCPs) were installed and point clouds were aligned in post-processing, but RTK allows immediate high-precision positioning on site, greatly reducing post-processing effort. As a result, the produced point clouds can be overlaid directly onto maps and design coordinate systems, alleviating accuracy management concerns.

Avoiding scan omissions through shooting plans and control point placement

To eliminate scan omissions, prior shooting and scanning planning is essential. Based on site geometry, determine “from which locations and in which directions should we scan to cover the whole area,” and combine overhead or oblique shots from elevated positions as needed. For TLS measurements, scan the building from multiple positions around its perimeter to cover blind spots and ensure sufficient overlap when merging point clouds. In photogrammetry, capturing oblique images and supplementing with ground-level shots for missing areas allows even the backside of a model to be digitized. Also, placing control points or target markers at key locations creates visible indicators that a spot has been measured and serves as references for later data integration. For instance, when scanning a long tunnel in sections, placing marks at regular intervals so each section includes them ensures the entire alignment can be connected without omissions. These measures minimize scan gaps even for complex structures.

Lightening and routinizing data capture with mobile LiDAR and smartphone point clouds

You no longer need large-scale equipment to capture point clouds; mobile LiDAR and smartphones can now easily capture point clouds. Using the LiDAR sensor on higher-end iPhones or iPads, you can immediately scan point clouds for ranges of several meters (several ft) with dedicated apps. While the number of points and accuracy are inferior to stationary lasers, the convenience of using a smartphone you carry every day to measure in spare moments is a major advantage. In addition, backpack and handheld LiDAR devices have recently appeared, allowing wide-area scanning while walking. These mobile scanners capture on the order of hundreds of thousands of points per second and stitch data together by estimating their own position using SLAM (simultaneous localization and mapping). Though not reaching the precision of high-end TLS, they can achieve accuracy on the order of several centimeters (approx. several inches), which is practical for daily construction progress recording and measurements in confined areas. Because heavy equipment is not required, the hurdle to point cloud measurement drops, and site workers themselves can routinely “take a quick scan,” enabling everyday use.

Improving shareability via cloud processing and point cloud viewers

Processing and sharing massive point cloud datasets can be drastically improved by using cloud services. Heavy computations such as generating point clouds from drone photos can be offloaded to high-performance servers in the cloud, delivering results quickly without relying on local PC specs. Services that automatically convert uploaded on-site photos into point clouds and orthophotos are now available. When sharing completed point clouds, a cloud point cloud viewer makes distribution simple. By publishing the data to the cloud and sending the generated URL, recipients can view and measure 3D point clouds in a browser. This approach removes the need for recipients to have dedicated software or high-performance PCs. For example, site supervisors or clients can open a link on an office laptop to display the latest point cloud in 3D and check details using distance measurement or cross-section tools. Multiple stakeholders can view the same platform simultaneously, reducing misunderstandings and smoothing consensus-building. Because data on the cloud is always the latest version, the risk of “referring to an outdated version” is also prevented.

Combining with LRTK to enable high-precision positioning and point cloud recording with a single smartphone

Finally, a cutting-edge solution is the integration of smartphones with RTK positioning. Pocket-sized RTK-GNSS devices that can be attached to smartphones—such as Japan-origin LRTK systems—have emerged, turning a smartphone into a centimeter-class surveying instrument. For example, attaching an LRTK receiver to an iPhone allows high-precision positional coordinates to be automatically and in real time assigned to point clouds scanned by the phone’s built-in LiDAR or camera. Previously, point clouds were adjusted based on separately surveyed reference points, but with LRTK the scanning and positioning are completed simultaneously on site, producing point cloud data that already conforms to geodetic coordinates without post-processing. Moreover, data can be uploaded from the smartphone to the cloud immediately, enabling office staff to check results in real time. The smartphone + LRTK combination is very lightweight—about 125g—and can be carried at all times, making it effective in situations such as disaster sites where large equipment cannot be brought in. Realizing “one-person-one-smartphone surveying” dramatically improves both the accuracy and mobility of point cloud scanning and supports DX (digital transformation) of field operations.

Summary

Point cloud scanning is a powerful means of digitally capturing the as-built site with millimeter- to cm-level accuracy (half-inch accuracy), but its adoption and use involve various challenges from accuracy management to data processing. This article listed issues such as coordinate alignment, equipment and data constraints, and operational difficulties, and proposed solutions for each. Technologies that address these challenges—such as using GNSS/RTK for absolute accuracy, cloud services for efficient data sharing, and the revolutionary workflow of smartphone + LRTK—have increasingly become practical. The key is to actively adopt these new tools while combining them with site-specific measures (advance planning, training systems, etc.) in operational practice. By fully leveraging the potential of point cloud scanning, surveying and construction management accuracy and productivity can be greatly improved. Easier detailed site understanding and sharing obtained data among stakeholders will further accelerate DX in future construction and civil engineering sites.