Drone × SfM Processing Opening the Way for Construction DX: Latest Cases of Rapid Point Cloud Generation

The construction industry urgently needs "Construction DX (digital transformation)" aimed at improving productivity and reducing labor. One trump card attracting attention is the combination of drone aerial photography and SfM processing (Structure from Motion) to rapidly generate 3D point cloud data. This article explains how SfM processing works, the advantages of drone-based wide-area rapid surveying, implementation examples such as autonomous flight, RTK, and the latest LRTK technology, and use cases in as-built verification and disaster recovery. It also introduces differential detection using point cloud data (heatmap visualization) and the benefits of GCP simplification and smartphone integration via LRTK, and concludes with an outlook on operational efficiency and labor reduction brought by these technologies.

1. What is SfM processing? Photogrammetry mechanism and 3D point cloud generation technology

SfM (Structure from Motion) processing is a photogrammetry technique that reconstructs the three-dimensional structure of an object or site from multiple photographic images. Using computer vision algorithms, it detects feature points in overlapping areas of multiple captured photos and matches those common points to automatically compute the camera positions and orientations and the 3D coordinates of each point. As a result, the shooting positions and attitudes for each photo are estimated, and a point cloud data set (3D point cloud) representing the object’s shape is generated. Because computers automate the aerial photogrammetry process that previously required artisanal skill, a major characteristic is that high-density 3D models can be efficiently constructed from many photographs.

The generated point cloud data represent surfaces of objects or terrain with countless points, each having X, Y, Z coordinate information (and color information based on the photos). The site can be recreated like a lifelike 3D photograph, and it is easy to analyze distances and volumes from the acquired point cloud or create cross-sectional drawings. In addition, point clouds derived from photos can produce texture-mapped 3D models, which can be overlaid with design data or visualized in VR. SfM processing can be run automatically using dedicated software (e.g., Metashape or Pix4D), so as long as you have images taken by a camera, high-accuracy 3D surveying has become accessible to anyone.

2. Why drone (UAV) aerial photography + SfM can acquire point clouds quickly and over wide areas

The combination of drone (UAV) aerial photography and SfM processing is a powerful means to acquire wide-area 3D point clouds in a short time. Shooting from the air allows coverage of a large area at once, making it far more efficient than ground crews surveying on foot. For example, a large-scale earthwork site that would previously take surveyors several days with total stations can be photographed by drone in just tens of minutes to half a day, producing hundreds of photos and a detailed point cloud model with centimeter-level accuracy (half-inch accuracy). In fact, the Ministry of Land, Infrastructure, Transport and Tourism reported a case where terrain surveying of about 0.3 square kilometers (0.1 sq mi) that previously took 45 days was shortened to 1.5 days using UAV laser surveying. Similar gains are seen in photogrammetry, where on-site work time can be dramatically reduced compared to manual survey point observation.

Another reason is that the high-density photographic data from drones and advances in SfM software have increased the speed of data processing (office work). By parallel processing on high-performance PCs or cloud services, point cloud generation from hundreds to thousands of photos can be completed in a short time. This enables rapid cycles of surveying → point cloud generation → analysis, making it realistic to scan sites weekly or even daily to monitor progress. Drones also allow safe aerial situation assessment of steep slopes or hazardous areas immediately after disasters where people cannot enter. Drone + SfM, which can digitally capture wide areas with high accuracy, is increasingly recognized as a foundational technology supporting digital twin initiatives for construction sites.

3. Implementation examples such as autonomous flight, RTK correction, and LRTK-equipped drones

Various technical innovations have been implemented recently to obtain precise point clouds efficiently with drones. First, autonomous flight (self-piloting) enables anyone to perform stable aerial photography. With dedicated flight planning software, specifying the survey area allows the drone to automatically fly a regular route and capture photos at optimal overlap levels (typically front overlap > 80% and side overlap > 60%). This yields uniform, high-quality photographic data, stabilizing the accuracy of point cloud generation by SfM. During autonomous flight, AI-based obstacle detection and altitude-hold functions operate, allowing safe and reliable wide-area aerial photography even for non-experts.

Next, high-precision positioning via RTK correction is indispensable. RTK (real-time kinematic) is a technique that augments satellite positioning like GPS in real time, enabling centimeter-level position information via a GNSS receiver mounted on the drone. Normally, drone photogrammetry optimizes photo positions in software afterward, but using RTK records high-precision geocoordinates for each photo, dramatically improving the model-wide georeferencing accuracy. Specifically, standalone positioning errors that used to be on the order of several meters are reduced to less than a few centimeters with RTK, so accurate 3D models can be obtained without pre-deploying many ground control points (GCPs). RTK corrections can be received by placing a base station on site and communicating via radio, or by connecting to network-type RTK (VRS) services provided by entities such as NTT or the Geospatial Information Authority of Japan. The latter allows correction information to be obtained via the Internet, enabling RTK positioning on site without dedicated radios or a base station.

Furthermore, a new initiative is the emergence of LRTK-equipped drones. LRTK (Lightweight RTK) is a compact integrated RTK-GNSS device & service developed by a startup from Tokyo Institute of Technology; originally designed to attach to smartphones, its compactness and versatility have raised expectations for drone surveying applications. By mounting a palm-sized receiver weighing a few hundred grams on the aircraft and connecting it to an onboard computer via Bluetooth or Wi-Fi, high-precision positioning can be easily realized. If RTK positioning, which traditionally required specialized surveying equipment and heavy devices, becomes manageable even on small drones, a highly mobile and cost-effective surveying system can be built. In practice, attempts have already begun to retrofit LRTK modules to RTK-unaware consumer drones to enable centimeter-level aerial photogrammetry on inexpensive platforms. The spread of the "precision autonomous drone surveying" combination of autonomous flight + RTK is expected to make wide-area point cloud measurement more accessible and highly accurate.

4. Use cases and outcomes of SfM in as-built verification, progress management, disaster recovery, and terrain surveying

Point cloud technology from drone aerial photography and SfM is already being applied across civil engineering and construction sites. Here are some representative use cases and the results achieved.

• As-built verification: Point clouds are powerful in verifying whether completed work matches the design. For example, in road construction, finished road surfaces and slopes are converted into point clouds via drone photogrammetry and overlaid with design 3D models or drawings for comparison. Because thickness and height differences can be analyzed across surfaces on the point cloud, judgments about as-built quality that were previously based on limited survey points can be evaluated across the entire area, improving quality control accuracy.

• Progress management: SfM point clouds are also useful for tracking construction progress. Regularly (e.g., weekly or monthly) updating site point clouds with drone flights allows quantitative grasp of volume changes from embankments or excavations. For example, point cloud comparisons can calculate how much earth has been removed or brought in since the last survey, helping objectively measure progress and manage schedules. There are also use cases where freshly placed concrete is scanned and compared with the design model to verify that the specified thickness and slopes are achieved immediately after placement. Point cloud–based progress management advances site visualization and automates recordkeeping, streamlining report generation.

• Disaster recovery: Drone + SfM is effective for assessing landslides and debris flows. Since detailed 3D surveying is possible from the air even in areas unsafe for people, rapid estimates of collapsed soil volumes and mapping of damage extent can be performed. In one earthquake-affected site, drones were used immediately after the event to capture the damaged area and create a point cloud model, enabling quick estimation of collapsed soil volumes to inform emergency restoration planning. Furthermore, by leveraging LRTK as mentioned later, workers could measure and photograph detailed points in the disaster area with smartphones and share the data in the cloud, allowing remote headquarters to instantly grasp the situation and make decisions. Situational assessment that previously took days can now be completed within hours in some cases, contributing to faster disaster response.

• Terrain surveying: SfM point clouds are used for terrain surveys needed during road or land development planning. Tasks that used to require survey teams to traverse mountains and measure many points can now produce a high-density terrain model in a short time by photographing from the air with drones. Even in rugged terrain, contour maps and longitudinal/cross sections can be generated freely from point clouds, enabling designers to plan with detailed topographic information. Especially for forest road construction and erosion-control works, acquiring up-to-date terrain point clouds in advance streamlines optimal route selection and earthwork quantity calculations. Based on such results, the government's "i-Construction" initiative recommends drone photogrammetry, and many cases report shortened surveying periods and cost reductions.

5. Differential detection (deviation from design, progress, volume) by point cloud comparison and heatmap visualization techniques

Point cloud data obtained by SfM can be compared with design models or past point clouds to derive various differential information. This allows quantitative evaluation of site progress and quality and intuitive identification of issues. Representative methods of differential detection and visualization include:

• Detection of deviations from the design model: Overlay the completed point cloud with the design 3D model (or design terrain data) and compute offsets at each location. In as-built verification, visualizing these offsets as a color distribution (heatmap) makes it immediately clear which areas are overfilled or undercut relative to the design elevation. For example, for earthwork foundation leveling, generating a map from the point cloud that marks areas more than 5 cm high in red and low areas in blue lets you quickly identify spots needing rework. Where previously finish errors were inferred from a very small number of inspection points, point cloud comparison enables comprehensive surface-wide detection.

• Calculation of progress and earthwork volumes: Comparing the latest point cloud with the previous survey allows accurate calculation of earth volume changes due to construction. For instance, in embankment works, you can verify whether the fill volume within a section matches the design by computing differences between before-and-after terrain point clouds. If the transported-out or transported-in soil volumes at a development site are automatically calculated from point cloud differences, the progress management figures gain credibility. Combined with heatmap display, you can show where and how much soil has increased or decreased by color, making it a useful, intuitive reporting material for progress meetings.

• Monitoring deformations and displacements: Comparing multiple point clouds taken at different times can detect minute displacements in structures or terrain. For example, in periodic bridge inspections, overlaying a point cloud from a year earlier with the current one and color-coding differences can detect pier settlement or deck deflection changes at the millimeter level. Long-term changes that human eyes might miss can be quantitatively captured by digital point cloud comparison. This technology contributes to advanced infrastructure maintenance and reduces inspection labor.

Results of these differential analyses can be output as numerical reports in software or shared as color heatmap diagrams. Visualization by point cloud comparison enables construction managers to early detect deviations from design or schedule delays and take corrective measures at the right time. Point cloud data truly act as a mirror reflecting the site’s "truth," driving the shift from intuition-based decisions to data-driven management and accelerating site DX.

6. Field benefits of LRTK correction observation, GCP simplification, and smartphone integration (complementing ground blind spots)

While RTK technology discussed earlier is important for efficiently obtaining high-precision point clouds, the recently introduced LRTK (lightweight RTK solution) is becoming a game changer. LRTK consists of a compact and lightweight RTK-GNSS receiver, a smartphone app, and cloud services, enabling centimeter-level positioning for anyone. By attaching a dedicated receiver to a smartphone and connecting to a network RTK correction service via the app, a single operator can immediately begin high-precision positioning. Compared to conventional RTK survey instruments, LRTK receivers are overwhelmingly compact (the receiver is about 10 cm (3.9 in) in diameter and weighs a few hundred grams), making it easy to carry on a survey pole while walking or mount on a helmet to free both hands for measurement—excellent for field handling. No complex equipment operation is required, and satellite count and positioning accuracy are displayed at a glance on the smartphone screen, so non-specialist technicians can quickly master its use with short training.

Introducing LRTK significantly simplifies the placement of GCPs (ground control points) required for photogrammetry. Even when flying an RTK-equipped drone, several control points are usually placed on site for accuracy verification or contingency, but using LRTK allows one person to measure those coordinates efficiently. Walking around with a smartphone, tapping a button at the center of markers or reference points instantly acquires high-precision coordinates and records them in the cloud. This reduces a baseline survey that used to take two or more people half a day to a short task, allowing immediate transition to aerial shooting. Moreover, even when performing photogrammetry with small drones that lack onboard RTK, preparing a few GCPs measured by LRTK can ensure accuracy, making inexpensive platforms practically useful.

In addition, LRTK’s integration with smartphone cameras is powerful for complementing ground blind spots. Areas not fully captured by drone aerial photography—such as the backside of structures, undersides of bridge girders, or under dense tree canopies—need ground photos added to the SfM processing. If those ground photos are taken with a smartphone (or tablet) equipped with LRTK, accurate position coordinates are assigned to each ground image, smoothing integration with aerial photos. For example, images of bridge undersides or complex piping around plant facilities taken from the ground that include LRTK-derived position information can be easily registered with other aerial images in SfM software. As a result, you can build a fully covered 3D model including areas that drones cannot capture. On site, supplementary photos can be shared to the cloud in real time and point cloud merging and analysis performed at headquarters immediately, enabling real-time coordination. LRTK makes data sharing between field and office seamless and speeds up decision-making.

Thus, LRTK supports site DX as a drone-survey–friendly "easy high-precision positioning tool." Major domestic construction companies and surveying firms are already conducting pilot introductions, and the concept of "one GPS surveying device per person" is expected to drive future adoption.

7. Conclusion: Operational efficiency and labor reduction brought by LRTK adoption

We have reviewed the latest cases and technological trends in point cloud generation using drone × SfM, and at the heart of this is RTK enabling real-time high-precision positioning and next-generation LRTK solutions. Utilizing these technologies dramatically streamlines processes from surveying to construction management, and labor reduction on site is becoming tangible. Autonomous drone flight makes data acquisition accessible to anyone, SfM processing automatically generates massive 3D information, and LRTK allows even single operators at sites with a shortage of specialists to perform precise surveying—digital technologies are transforming construction sites and powerfully driving Construction DX.

As smart positioning technologies like LRTK spread, the era in which field and office are directly connected by data will take hold. With point cloud data shared in the cloud, remote offices can grasp current conditions and issue instructions, AI analysis can automate as-built checks, and new workflows will emerge. This means limited personnel can safely and accurately manage many sites, which will be a great boon to the construction industry struggling with severe labor shortages.

The effects of introducing LRTK extend beyond mere improvements in surveying accuracy. Expect ripple benefits such as shortened lead times for data acquisition and processing, reduced heavy labor, simplified reporting tasks, and—most importantly—faster decision-making by closing the information gap between field and office. These advantages naturally lead to on-site labor reduction and contribute to work-style reform and improved safety.

Finally, the combination of drone × SfM × LRTK is a powerful solution that materializes DX at construction sites. By establishing a system that can rapidly acquire and utilize high-precision 3D point clouds, site operations will surely advance to the next stage. If your site has not yet adopted these technologies, don’t miss this wave—consider experiencing smart surveying once. A single step toward new technology could significantly boost tomorrow’s site productivity and competitiveness.