SfM Processing Complete Guide: Industry Case Studies, Benefit Comparison, and Thorough Explanation from Tool Selection to Implementation

What is SfM Processing? Basic Concepts and Reasons for Attention

SfM (Structure from Motion) processing is a computer vision technology that reconstructs the three-dimensional structure (3D model) of an object or site from multiple photographic images. Also known as photogrammetry, it automatically analyzes images captured from various angles by drones or cameras to generate point cloud data, 3D mesh models, orthophoto (ortho) images, and more. Traditionally, 3D measurement required expensive laser scanners (LiDAR), but with the spread of SfM, ordinary digital cameras and drones can now easily acquire high-density three-dimensional data.

The reason SfM processing has attracted attention recently is its ease of use and cost benefits. Without relying on specialized equipment or expert technicians, site personnel themselves can capture wide areas in a short time and automatically create 3D models, dramatically improving work efficiency and productivity. In addition, the acquired data can be stored as digital point clouds and models, enabling intuitive site understanding and advanced process management that were difficult with traditional 2D drawings or photos. Governments are also promoting ICT utilization as part of productivity improvement measures in the construction industry (e.g., *i-Construction*), and 3D surveying using drone aerial photography and SfM is becoming an industry standard. SfM processing, which provides low-cost yet high-accuracy 3D data, is expected to be a key technology supporting modern digital transformation.

Industry Use Cases

SfM processing is already being practically applied across a wide range of fields. Below are representative use cases by industry.

• Construction sites: In civil engineering and construction, aerial SfM surveys using drones enable rapid acquisition of terrain data over wide areas, which is used for earthwork planning and as-built management. Tasks that traditionally took several days—such as site surveying and earthwork volume calculation—have been greatly streamlined by drone imaging and automated processing. Comparing generated 3D models with planned data allows early detection of construction errors and visualization of progress. Another advantage is the ability to safely survey hazardous slopes and embankments without having personnel enter dangerous areas.

• Municipal infrastructure inspection: Local governments use SfM-based 3D models for maintenance and management of infrastructure such as bridges, tunnels, roads, and dams. High-resolution models of structures captured by drones help identify cracks and displacements and quantitatively record deterioration over time. Digitalizing inspections that inspectors once performed visually enables the formulation of repair plans based on objective data. Especially for high or extensive facilities, SfM reduces the cost and risk of erecting scaffolding or using aerial work platforms, contributing to greater efficiency and safety in municipal operations.

• Disaster prevention: SfM is also a powerful tool in disaster response. By aerially photographing sites affected by landslides or flooding and instantly creating 3D terrain models of the affected areas, responders can grasp damage and assess secondary disaster risks. For example, the volume of a collapsed slope can be measured from the model to quickly estimate the amount of debris to be removed. In peacetime, terrain models are used to create debris-flow hazard maps or to periodically monitor deformation of levees and slopes, demonstrating how SfM data supports preventive disaster measures. The ability to safely obtain detailed data from the air in dangerous areas where people cannot approach greatly aids disaster response operations.

• Cultural heritage preservation: In the field of historical buildings and cultural assets, 3D recording via SfM is becoming an indispensable method for preservation and restoration. Sites such as ruins, temples, and shrines can be precisely scanned without contact and archived digitally for future restoration work and research. Damage that is difficult to see with the naked eye can be inspected in detail on the model, allowing a complete record of the current state. Because detailed dimensional information can be obtained without moving or touching the actual artifact, high-accuracy drawings and restoration planning can be carried out while protecting valuable cultural properties. The generated 3D models can also be used in VR content and online exhibitions, contributing to the digital publication and tourism promotion of cultural assets.

• Agriculture: Drone image SfM processing is also used in agriculture. 3D maps generated from aerial images of fields (hojō: farmland) help identify low spots with poor drainage and measure field slopes, aiding drainage measures and land leveling. In orchards, modeling tree height and canopy volume supports monitoring growth and yield prediction. Visualizing entire agricultural areas makes it possible to detect growth variability and early signs of disease that were previously overlooked, contributing to precision agriculture (smart farming). Covering surveys that relied on manual labor with aerial imaging also helps reduce workload, and this application is expected to expand further.

Comparison of SfM Implementation Benefits (Cost, Accuracy, Operational Burden, Compatibility with Digitalization)

The benefits of introducing SfM technology are organized here in comparison with conventional methods.

• Cost: SfM can greatly reduce initial investment and operating costs compared to traditional surveying and measurement methods. Dedicated 3D measurement equipment like laser scanners used to cost millions of yen, but with SfM you can start with a commercial drone or camera and a PC. Fewer personnel are required for shooting, and a single operator can cover wide areas, leading to labor cost reductions. Software options range from open source to cloud services, allowing flexible adoption according to scale. For these reasons, low-cost 3D adoption is achievable even for small- to mid-scale projects, which is a major advantage.

• Accuracy: While some worry about the accuracy of photogrammetry, current SfM technology can achieve accuracy on the order of a few cm (a few in) if proper procedures are followed. In particular, using an RTK-GNSS-equipped drone for imaging or placing known coordinate control points on the ground provides absolute accuracy (positional accuracy) to the model. This allows the resulting point clouds to be overlaid directly on public coordinate maps or GIS. On the other hand, SfM has weaknesses compared to LiDAR, such as data gaps in areas difficult to capture in photos, like ground under tree canopies. However, many issues can be mitigated by planned shooting or supplementary measurements, and for typical civil surveying and design purposes, SfM can secure sufficient accuracy.

• Operational burden: The operational hurdles for SfM implementation have also decreased compared to the past. On-site shooting is faster and simpler than detailed traditional ground surveying. While drone setup and flight planning are required before takeoff, once airborne the drone can automatically capture photos, reducing the physical burden and risk to operators. Post-processing used to require high-performance PCs and long computation times due to large-scale image processing, but nowadays cloud services allow processing without relying on the field PC’s specs. Software automation and streamlined workflows mean that even non-specialist technicians can handle SfM with relatively short training. Overall, SfM has reached a stage where it can be integrated into daily operations without undue strain, and operational burdens are steadily decreasing.

• Compatibility with digitalization: SfM outputs are digital by nature, making integration with other ICT tools easy. For example, point clouds and 3D models can be imported into CAD or BIM/CIM software for cross-checking with design data, enabling intuitive interference checks and as-built verification that are difficult with traditional 2D drawings. Orthophotos and point clouds can also be overlaid on GIS maps to manage infrastructure asset information. Sharing via the cloud allows remote stakeholders to view current site conditions in 3D, improving information sharing and decision-making speed. As organizations shift from paper-centric to digital-centric workflows, SfM processing strongly supports such DX (digital transformation).

Tool and Software Selection Points (Beginner-Friendly, Industry-Specific, Cloud-Based, etc.)

Many software and services are available for SfM processing. Here are points to consider when selecting tools that match your organization’s use cases and skill level.

• Beginner-friendly: For first-time SfM adopters, choose tools with user-friendly operation. Software with intuitive interfaces and comprehensive Japanese support or manuals facilitates smooth learning. Some tools automatically process photos simply by drag-and-drop, making them usable without specialized knowledge. It’s important to try free or trial versions to assess usability and confirm whether site personnel can handle the tool.

• Industry-specific: There are purpose-specific solutions among SfM processing tools. For example, civil engineering-focused tools may have robust volume calculation and terrain mapping functions, while cultural heritage-focused tools may excel at high-resolution textures and fine-detail modeling. In agriculture, services with GIS integration and growth analysis features may be preferable. Compare whether a tool provides functions suited to your use cases and which tools are widely adopted in other companies to avoid mismatches.

• Cloud-based: In addition to traditional desktop-installed software, cloud-based services that complete processing online are increasingly available. Advantages of cloud-based solutions include fast processing of large photo sets regardless of local PC specs and no need for installation or updates. Results are easy to share and view via the web, allowing model checks on a unified platform across remote field offices. However, uploading large numbers of photos requires adequate network connectivity, and running costs such as monthly fees should be considered. Evaluate cloud suitability against project frequency and data confidentiality.

• Other considerations: Beyond the above, implementation cost and budget are important. Free open-source tools reduce initial costs but often rely on self-support. Paid software offers strong support and rich features but requires balancing license costs. Also confirm that output data formats are compatible with your workflows (e.g., LAS point clouds, OBJ models, GeoTIFF orthos). Consider your organization’s IT literacy and operational structure and select tools from the perspective of long-term usability, which is key to success.

Precautions at Implementation and On-Site Checklist for Utilization (Photo Shooting, Ensuring Accuracy, Processing Load, etc.)

When newly introducing SfM processing on site, there are key points and on-site precautions to keep in mind. The following checklist summarizes them.

• Photo shooting basics: The quality of 3D reconstruction heavily depends on the source photographs. Ensure sufficient overlap rate (60–80% or more) and capture subjects from various angles. Out-of-focus or blurred images degrade accuracy, so adjust camera settings (appropriate shutter speed and ISO sensitivity) to obtain sharp images. For outdoor shooting, pay attention to sunlight conditions: avoid extreme backlighting and aim for uniform lighting to stabilize post-processing.

• Measures to ensure accuracy: Depending on the required accuracy level, additional shooting measures may be needed. To improve absolute accuracy, place known coordinate control points at the site and include them in the photos. These points can be used later to scale and georeference the model, yielding more accurate dimensions and coordinates. If the drone is equipped with high-precision GNSS, the photo positions along the flight path can be positioned, simplifying the need for control points. In any case, consider appropriate positioning aids according to the necessary accuracy.

• Processing environment and data management: After shooting, large numbers of photos must be processed, so confirm PC performance or cloud service use in advance. For local PC processing, handling several hundred photos or more requires a high-performance GPU and sufficient memory. If using cloud processing, schedule extra time to account for upload duration. Generated point clouds and models can be very large, so prepare storage capacity and a backup system for file management.

• On-site data checks: Before leaving the site, make it routine to confirm that there are no missing or defective data. For example, quickly check the captured images on site for blur or missed areas (areas not fully covered). If necessary, perform additional shooting on site to prevent later issues such as “the model has holes” or “important parts lack sufficient accuracy.” If possible, process a subset of photos on a laptop to get a rough idea of quality before leaving.

• Compliance with regulations and safety measures: When using drones, flight plans must comply with aviation law and related regulations. Confirm no-fly zones and altitude limits and obtain necessary permits in advance. Securing safety around the site is also important. Assign watchers where third parties might enter the area to prevent accidents. Do not neglect basic checks such as battery levels and aircraft inspection—safety first ensures smooth SfM utilization.

Conclusion: Integration with LRTK for Simple Surveying and Implementation Effects

SfM processing brings the various benefits introduced above through photogrammetry using drones and cameras, and recently its effectiveness has been further enhanced by combining it with other advanced technologies. One example is integration with the simple surveying solutions provided by LRTK.

For example, using the smartphone-integrated surveying device "LRTK Phone" enables anyone to easily obtain positioning points with cm-level accuracy (half-inch accuracy) or perform handheld detailed scanning. Overlaying precise point cloud data acquired with the LRTK Phone onto point clouds generated by drone SfM creates a seamless 3D model that includes areas that drones cannot see (such as under tree canopies or behind structures). By combining multiple measurement methods, all site information can be fully digitized.

Adopting such an integrated approach results in dramatic productivity improvements in surveying work. Tasks that were previously conducted by separate teams and equipment become centralized, and the workflow from field work to data processing and sharing flows smoothly, directly shortening project lead times and reducing costs. The higher reliability of obtained data also makes it easier to share the "truth of the site" among stakeholders, enabling optimized construction planning and faster decision-making. In practice, construction companies and municipalities have begun combining drone SfM and LRTK surveying, achieving significant results in initial disaster response surveys and infrastructure management.

Finally, digital surveying centered on SfM processing will increasingly become a routine work tool. By adopting these technologies in ways that fit your company’s needs, you can steadily advance on-site DX and realize operational reform that balances safety, efficiency, and accuracy. Use this guide as a reference to consider implementing SfM processing and integrating it with LRTK simple surveying at your company.