How SfM Processing Is Revolutionizing Construction Surveying: Generating Precise 3D Point Clouds from Photos

In recent years, "SfM processing" for generating 3D point clouds from photographs has attracted significant attention at construction and civil engineering surveying sites. What once required expensive surveying equipment and considerable effort—3D measurement of as-built conditions—can now be performed easily with consumer digital cameras or drones, dramatically improving both the efficiency and accuracy of surveying.

This article explains the mechanisms and methods of SfM processing in an easy-to-understand manner, explores the differences and advantages compared to traditional surveying methods, and presents concrete applications in construction surveying such as as-built management, earthwork volume calculations, and pre-construction terrain surveys. It also introduces recent cases combining drones, smartphones, and cloud services. In addition, we discuss the limitations of SfM and comparisons with complementary technologies such as GNSS and LiDAR, as well as challenges and points of caution when implementing SfM. Finally, we summarize effective methods for correcting the accuracy of point cloud data obtained by SfM and unifying coordinate references using the simple survey approach LRTK with a smartphone plus a high-precision GNSS receiver.

What SfM Processing Is (Mechanism and Methods)

SfM (Structure from Motion) is a technique for reconstructing the three-dimensional structure (3D model) of an object or site from multiple photographic images. Originally a technique in the computer vision field, it has recently been widely applied to photogrammetry in civil engineering and construction.

In SfM processing, feature points (high-contrast patterns, corners, etc.) are automatically detected across multiple overlapping photos, and matching these common points allows simultaneous estimation of the camera positions and orientations at capture time and the 3D coordinates of each feature point. Based on this estimate, image information from multiple viewpoints is integrated to generate a high-density point cloud (detailed reconstruction via Multi-View Stereo). This process enables precise reproduction of site topography and structures as detailed 3D point clouds or polygon mesh models.

The basic workflow of SfM processing is as follows:

• Capture numerous photos of the site from various angles (ensure sufficient overlap between adjacent photos).

• Import the captured photo data into specialized software and automatically detect feature points in each image.

• Match corresponding feature points between photos and compute camera positions and orientations (photo alignment).

• Analyze parallax from the estimated camera positions and reconstruct a high-density 3D point cloud (point cloud generation).

• If needed, create a 3D mesh model or an orthophoto (an image viewed from directly above) from the point cloud.

Thus, a key characteristic of SfM processing is that detailed three-dimensional data can be obtained by software processing from photos taken with standard digital cameras or drones, without special equipment.

Differences from Traditional Methods and Advantages (Accuracy, Efficiency, Cost)

With the advent of SfM processing, the following major differences and advantages have emerged compared to traditional surveying methods:

• Accuracy: With appropriate capture and processing, photogrammetry via SfM can produce 3D models with centimeter-level accuracy. This is sufficient for typical civil engineering as-built management and earthwork volume calculations. In single-point absolute accuracy, traditional total station surveys or high-precision GNSS positioning (RTK) can outperform SfM at the millimeter level in some cases. However, SfM can measure broad areas in a surface-wise manner, giving very high data consistency and overall accuracy; by referencing known points (ground control points) for correction, SfM can achieve accuracy comparable to traditional methods when required.

• Operational efficiency and safety: Photogrammetry using SfM brings dramatic efficiency improvements. For example, flying a drone can complete aerial photography of a large development site in tens of minutes and produce millions of survey points in a single flight. Compared to manual point-by-point surveying, on-site work time can be drastically reduced. Surveyors no longer need to enter hazardous slopes or roadways, significantly improving safety. Because measurements can be made later at a desk from the captured photos, repeat trips to the site for additional measurements are reduced, increasing overall workflow efficiency.

• Cost: Required equipment can be as simple as a camera or drone and a general-purpose PC, which lowers initial investment compared to purchasing and maintaining expensive equipment like 3D laser scanners (LiDAR). Reducing outsourced surveying and completing data creation in-house can achieve long-term cost savings. Storing detailed digital records of site conditions allows immediate reuse for design changes or additional work, minimizing rework and extra surveying costs. SfM, which can deliver high-density point clouds at low cost, is a powerful solution for sites with limited budgets or personnel.

Applications in Construction Surveying (As-Built Management, Earthwork Calculation, Pre-Construction Surveying, etc.)

High-precision 3D point cloud data produced by SfM can be used in various surveying and measurement tasks in construction. Major applications include the following:

• As-built management: You can record and verify the completed shape of roads or development sites in 3D. Overlaying point cloud data with the design model makes it easy to check at a glance whether embankment and cut/fill slopes and heights match the design, helping to identify nonconforming areas early. Traditionally, construction verification was done by sampling measurement points and checking cross-sections; point clouds enable surface-level and intuitive verification.

• Earthwork volume calculation: You can accurately calculate the volumes of excavation and embankment. For example, by creating point clouds of the ground surface before and after excavation with SfM and comparing them, cut-and-fill volumes can be computed with high accuracy. Point clouds reflect undulations that might be missed by traditional cross-section surveys, improving the accuracy of quantity estimation. Drone photogrammetry is also used to measure the volumes of material stockpiles on site.

• Pre-construction surveys: Prior to commencing construction, detailed records of current terrain and surrounding environment are useful. Aerial photography by drone over a wide area can preserve the locations of property boundaries, existing structures, and tree coverage as a three-dimensional terrain map. This helps planners grasp site conditions in detail during planning, aiding design changes and risk assessment. Preserving a pre-construction terrain model also enables post-construction comparisons for environmental impact assessments.

• Progress monitoring: Regular photogrammetry of the site aligned with construction progress can be used to visualize progress in 3D. For example, weekly drone flights with updated point clouds allow time-series tracking of earthwork progress and structure assembly. Because the entire site can be viewed from above, delays or mistakes can be identified and corrected early, and the data can serve as material for safety management.

These applications align well with the Ministry of Land, Infrastructure, Transport and Tourism’s i-Construction initiative and CIM (Construction Information Modeling), and 3D point cloud-based site measurement is becoming standardized in many public works projects.

Examples of Integration with Drones, Smartphones, and the Cloud

SfM processing becomes more efficient and convenient when combined with various devices and services. Here are examples of integration with drones, smartphones, and cloud services.

• Point cloud generation from drone aerial photography: Drones are ideal for acquiring aerial photos. Capturing nadir (top-down) photos from above enables rapid creation of wide-area terrain models and orthophotos. Especially when using RTK-equipped drones, position coordinates are recorded with high accuracy, so the resulting point cloud data can be output already aligned to the survey coordinate system. This reduces the need to install many control targets, and only a few points are required for accuracy verification. Drone photogrammetry is widely used for site terrain mapping and infrastructure inspection, and it is becoming a standard method on i-Construction sites.

• 3D measurement using smartphones: With improvements in modern smartphone cameras, everyday smartphones can conveniently capture photos for SfM. For small structures or indoor spaces, you can create 3D models simply by photographing from all sides with a smartphone. Additionally, recent iPhones and iPads include built-in LiDAR sensors, enabling dedicated apps to capture simple point clouds in real time. Combining a smartphone with a high-precision GNSS receiver (as with the LRTK approach described later) allows precise positioning information to be embedded in photos for SfM processing. On site, smartphones can complement drone captures by photographing areas difficult for drones to reach (under bridges, under trees, etc.), and these ground photos can be merged with drone-derived point clouds. Smartphone + SfM is cost-effective and easy to use, making it suitable for small-to-medium sites and urgent surveys.

• Combination with cloud services: SfM processing involves heavy computation with many images, but cloud-based point cloud generation services have become more prevalent. By uploading photo data via the web, these services automatically generate 3D point clouds and orthophotos, eliminating the need for a high-spec PC. Point clouds generated in the cloud can be shared with stakeholders over the internet and viewed/measured in browser-based 3D viewers. If you capture photos on site and immediately process them in the cloud, you may be able to review the finished point cloud model by the time you return to the office—enabling rapid utilization. Even without in-house specialists, cloud services provide access to advanced SfM analysis and lower the technical barrier to adoption.

Limits of SfM and Complementary Technologies (Comparison with GNSS and LiDAR)

Despite its convenience, SfM processing has cases where application is difficult and technical limitations. Therefore, GNSS positioning and LiDAR (laser scanning) are often combined to complement SfM’s weaknesses.

• Limitations of photogrammetry: Because SfM depends on optical images, it struggles to produce adequate point clouds on featureless surfaces or highly reflective materials (e.g., plain white walls, water surfaces, mirrors, nighttime photography). In areas densely covered by trees, aerial photos may not capture the ground surface well, resulting in gaps in the point cloud. Parts not visible in images (the back sides of objects or shadowed areas) cannot be captured, and holes may remain in the model even with multi-directional capture. Also, SfM-derived 3D models are fundamentally in a relative coordinate system, so scale (dimensions) and absolute position require some reference information. Drone photogrammetry is also weather-dependent; flying in strong winds or photographing in rain is difficult. In short, SfM alone cannot handle all situations.

• Correcting scale and coordinates with GNSS: Combining GNSS positioning with SfM resolves the aforementioned coordinate reference issues. If high-precision GNSS (RTK or PPK) is used to obtain camera capture positions or coordinates of control points, the SfM-generated point cloud model can be scaled and aligned to those known coordinates. This allows the 3D point cloud to have absolute position information in national geodetic or project coordinate systems. Verifying errors with control points also secures the reliability of the SfM model. For example, installing a few RTK-GNSS survey points on site can allow integration of SfM results with existing survey coordinates to centimeter-level accuracy—achieving precision management that SfM alone cannot provide.

• Comparison and complementary use with LiDAR: LiDAR (Light Detection and Ranging) measures distances to targets directly with laser light to produce point clouds. LiDAR surveys can operate at night or in dark environments, capture the shape of textureless objects, and handle multiple returns to measure ground beneath trees. LiDAR can achieve very high accuracy—high-end instruments can measure at millimeter precision. However, LiDAR equipment is expensive and requires specialized handling, making it less favorable in cost and convenience compared with widely available camera-based SfM. LiDAR point clouds are typically monochromatic collections of points and do not provide the color information or the clear orthophotos that photographs can. Therefore, in practice, hybrid approaches are increasingly common: use SfM for most measurements because it is convenient, and supplement with LiDAR for areas SfM cannot capture. Such hybrid surveying builds digital models that balance cost, accuracy, and completeness.

Challenges and Points of Caution When Introducing SfM (Accuracy Verification, Shooting Conditions, Point Cloud Processing)

When introducing SfM photogrammetry on site, several issues should be noted. Below are summarized methods for accuracy verification, shooting condition tips, and point cloud processing points.

• Accuracy verification: When using SfM for the first time or for critical measurements, verifying accuracy by combining it with traditional methods is essential. For example, survey a few points on site with a total station or RTK-GNSS and compare the coordinates of the same points on the SfM-derived point cloud to quantitatively evaluate errors. It is also important to check for scale errors by comparing reproduced reference scales in the software (e.g., a known-length measuring rod or grid) with actual measurements. Through these verification steps, confirm that SfM results meet the required accuracy; if not, take additional photos or add control points to correct errors.

• Key points for shooting conditions: Capturing high-quality photos is key to SfM accuracy. Ensure sufficient overlap (70% or more between adjacent photos) and photograph the subject from various angles. To obtain sharp images, be mindful of camera shake and subject motion, and use a tripod or a drone’s auto-flight functions when necessary. Shoot during bright daytime hours and avoid extreme backlighting or strong shadows to improve feature extraction accuracy. For large sites, divide the area and photograph it systematically to avoid omissions. Keeping focal lengths and resolutions consistent stabilizes later analysis. It is advisable to take a few test shots and perform a quality check in the software to confirm there are no reflections or missing areas before the main capture.

• Point cloud processing and data utilization: Point clouds generated from photos can become extremely large (tens of millions of points or more), so proper processing and management are necessary. Use filtering functions to remove isolated noise points or obviously mislocated points. If ground surface measurement is the goal, remove unnecessary objects such as trees or vehicles, or perform automated ground extraction (classification) to isolate the terrain. When raw point clouds are unwieldy, convert them to mesh models, generate contour lines, or extract required cross-sections as deliverables. Because point cloud file sizes can be large, consider cloud storage and dedicated viewers for sharing. Developing in-house personnel skilled in point cloud processing and conducting trials to gain operational familiarity are important in the early stages of introduction. Considering these points will help smoothly apply SfM photogrammetry to on-site operations.

Conclusion: Using LRTK (Smartphone + High-Precision GNSS) to Correct SfM Accuracy and Unify Coordinate References

SfM processing, which can generate precise 3D point clouds from photos, has brought major innovation to construction surveying. As described in this article, SfM alone is extremely useful for improving on-site measurement efficiency, but its value is further enhanced by combining it with other technologies.

Particularly noteworthy is the use of LRTK, which integrates a smartphone with a high-precision GNSS receiver. LRTK is a small RTK-GNSS receiver that can be attached to a smartphone, enabling centimeter-level positioning on site with ease. Using LRTK during photo capture can improve the positioning information recorded in each photo, giving the SfM-generated point cloud a definitive scale and coordinate reference. This greatly simplifies coordinate alignment that previously relied on control targets and known-point surveys.

Moreover, LRTK allows supplementary surveying of fine detail points not captured in the SfM point cloud and enables immediate measurement of verification points for as-built drawings—providing real-time on-site feedback. Because the smartphone + GNSS combination is simple and portable, one person can both capture photos and perform positioning in the field, enabling efficient surveying with limited staff.

Combining SfM photogrammetry with LRTK positioning realizes an ideal workflow of "rapidly 3D digitizing wide areas while placing the entire dataset onto accurate coordinates." In future construction sites, this kind of measurement DX centered on smartphones will continue to advance. By pairing innovative SfM techniques with practical tools like LRTK, you can improve the productivity of your everyday surveying operations.