Field surveying is being reinvented with smartphones and photo analysis technologies! Traditional surveying required specialized equipment and a lot of effort, but recently combining smartphones with SfM processing (photogrammetry) has made it possible for anyone to perform high‑precision 3D measurements easily. This article explains the benefits and latest trends of field surveying using SfM processing and introduces next‑generation surveying methods that can be completed using only a smartphone. Using the solution called LRTK mentioned in the title as an example, we explore the concrete picture of how field surveying is being updated.

Challenges in field surveying and the need for digitalization

In construction and civil engineering, it is essential to accurately measure and record the shapes of constructed structures and terrain through "field surveying" and "as‑built management." However, conventional surveying methods had several challenges. For example, they required a surveyor who could operate specialized instruments like total stations and levels, which involved significant personnel and time. On large sites, many survey points have to be measured one by one, creating a heavy workload. Also, because measurements were taken point by point, it was difficult to grasp the whole picture. Even if parts of the site differed from the design outside the measured points, there was a risk of overlooking them, sometimes leading to problems discovered later.

To solve these issues, digitalization of field operations (field DX) has been advancing, driven in part by initiatives like the Ministry of Land, Infrastructure, Transport and Tourism’s *i‑Construction*. One rapidly spreading approach is the use of 3D measurement technologies. Laser scanners and drone photogrammetry can capture an entire site’s shape as point cloud data at once, enabling planar and volumetric understanding and becoming a new standard. However, introducing expensive 3D laser scanners has cost and expertise barriers. Attention has therefore turned to more accessible SfM photogrammetry and the use of smartphones.

New possibilities for field surveying opened by SfM processing

SfM processing (Structure from Motion photo analysis) is a technique that reconstructs three‑dimensional shapes from multiple photos. For example, if you take photos of a site with a smartphone or digital camera and process that photo set with dedicated software, a 3D model of the entire site (point cloud or mesh) is generated. This allows you to obtain a high‑density digital point cloud that covers the whole site, rather than only a collection of manually obtained points. With photogrammetry using SfM processing, you can record large areas in a short time and later flexibly measure dimensions at arbitrary locations or extract cross sections for analysis.

Moreover, SfM photogrammetry can be practiced with everyday equipment. If you have a smartphone with a high‑resolution camera, you can scan a site without special surveying instruments. Although many cases previously relied on drone aerial photos processed with SfM, walking the site and photographing with a smartphone often yields sufficiently detailed data. Because smartphones are easy for anyone to handle, a major advantage is that site personnel themselves can take measurements without a specialist operator.

Why easy, high‑accuracy 3D measurement with a smartphone became possible

The combination of smartphones and SfM processing enabling easy yet high‑accuracy 3D measurement is due to technological advances. Recent smartphones have dramatically improved camera performance and can capture images with tens of millions of pixels, directly enhancing SfM accuracy. Smartphones also now include built‑in IMUs and altimeters, and some models support high‑precision GNSS, making it easier to utilize position and attitude information in analysis.

In addition, improvements in computing power for processing large numbers of images and the maturity of cloud services are important factors. If you upload data taken on a smartphone to the cloud, services that automatically perform heavy SfM processing on servers are now available. This enables you to obtain analysis results on site without carrying a specialist PC.

A decisive development was the emergence of solutions that combine smartphones with RTK‑GNSS (real‑time kinematic GPS). RTK enables centimeter‑level position accuracy for photo capture locations and feature point coordinates on the model, providing SfM‑derived 3D models with accurate scale and absolute coordinates. Traditionally, photogrammetry required measuring a reference length or placing ground control points to scale the model to real dimensions. RTK‑enabled smartphone measurement greatly reduces that effort, achieving both convenience and improved accuracy.

Practical field use of smartphone surveying realized by LRTK

Against this backdrop of technological evolution, the versatile surveying app known as "LRTK" has appeared. LRTK is a system provided for field use by Refexia Co., which attaches an ultra‑compact RTK‑GNSS receiver to a smartphone (iPhone/iPad). With LRTK, a single operator can perform surveying, point cloud measurement, and photography all in one using only a smartphone. Tasks that used to be split between surveying teams and photo‑documentation teams can be completed with a single device, greatly improving efficiency.

Particularly noteworthy is the photogrammetry function included in the latest LRTK. This implements SfM processing on a smartphone, allowing photos taken on site to be converted into 3D models immediately. LRTK links high‑precision RTK position information with SfM algorithms, so each point in the generated 3D model is assigned absolute coordinates (latitude, longitude, elevation) from global positioning. Therefore, viewing the finished model allows digital measurement of dimensions, slopes, areas, and volumes as if you were on site. Because the model already aligns with the geodetic coordinate system, it is easy to overlay the completed point cloud or 3D model onto BIM/CIM design data or GIS maps for use.

LRTK’s photogrammetry reportedly achieves accuracy around ±5 cm, delivering absolute precision that has been difficult to realize with ordinary photogrammetry. Moreover, since this processing can be completed on the smartphone itself, model generation is possible on the spot even in remote areas with unstable internet. Eliminating the time lag of taking data back to the office for PC analysis enables immediate on‑site verification, which is revolutionary.

Benefits of combining SfM processing with RTK positioning

Let’s summarize the benefits of combining SfM processing with RTK positioning for field surveying.

• Greatly expanded measurement coverage: Instead of point‑by‑point surveying, you can capture the entire site shape, enabling later analysis of arbitrary locations. This reduces missed areas and lowers the number of return visits for additional measurements.

• Improved efficiency and cost: With a single smartphone, one person can survey, reducing labor and equipment costs. There is less time spent moving and setting up, allowing wide areas to be covered quickly.

• High‑precision data acquisition: Using RTK‑GNSS gives acquired data absolute coordinates, so the resulting 3D models have accurate scale and conform to public reference coordinate systems. This facilitates comparisons with design drawings and other survey data in later stages.

• Improved safety: Dangerous slopes and high places can be documented from a distance by photography, reducing the need for workers to enter hazardous areas and contributing to overall site safety.

• Accumulation of digital records: If point clouds and 3D models are stored in the cloud, you can review site conditions later. This is useful for sharing and verifying construction progress and as evidence of completed work.

In these ways, smartphone surveying that merges SfM processing with RTK positioning is an innovative approach that brings benefits to quality, efficiency, and safety.

Conclusion: Next‑generation field surveying with LRTK

Photogrammetry with SfM, smartphones, and RTK positioning—combining these three approaches is now updating field surveying. Surveying that used to be the domain of specialists is becoming a digital tool accessible to everyone on site. A representative example is smartphone surveying solutions like LRTK, which are already attracting attention as tools that accelerate field DX.

If you feel limitations or inefficiencies with traditional surveying methods, consider simplified surveying with LRTK. This system enables high‑precision 3D measurement with familiar devices and is designed to be easy for first‑time users of digital photogrammetry. Introducing next‑generation field surveying using the latest technologies can simultaneously improve efficiency and accuracy. With LRTK and SfM processing as the new standard for field surveying, bring a wave of DX to your site.

Beginner’s guide to introducing SfM processing: comprehensive support from preparation to 3D model creation

What is SfM processing? What preparation is needed, and how do you create a 3D model from photos? This article thoroughly explains the basic concepts of SfM (Structure from Motion), common pitfalls for beginners, required equipment and preparation, photography tips, and the actual processing steps. Even those trying SfM for the first time will understand the flow from preparation to completed 3D model after reading this.

What is SfM processing? The technique that generates 3D models from photos

SfM processing reconstructs three‑dimensional models of objects or sites from multiple photographic images. SfM stands for "Structure from Motion," meaning it estimates structure from camera motion. You take many photos from various angles with drones, digital cameras, or smartphones, and by analyzing the displacement of feature points across images, the three‑dimensional shape of the subject is computed. The resulting 3D data can be point clouds or polygon meshes, and you can generate textured 3D models with photo colors projected onto surfaces or orthophotos viewed from above. In short, using SfM processing, you can produce precise 3D models of real objects with just commonly available cameras—no special 3D scanner required.

Equipment and preparation needed for SfM processing

For beginners starting SfM processing, first confirm the necessary equipment and preparation. Basically, expensive specialized equipment is not required, and you can start with items at hand.

• Camera or smartphone: Prepare a device capable of taking high‑quality photos. Typical digital cameras or modern smartphones are sufficient. Cameras with more megapixels and better lens performance will capture finer details.

• PC: You need a computer to process the photos. SfM processing analyzes many images, so a relatively high‑performance PC (with ample memory and GPU) is desirable, but if you use fewer photos, a standard‑spec machine can suffice.

• SfM software: Prepare dedicated software to generate 3D models from photos. There are commercial and open‑source options; beginners should choose software with an easy interface and good tutorials. Some software offers a GUI for simple processing, while others allow fine parameter adjustments.

• Target markers or scale (as needed): If you want the model to reflect real‑world scale, include a scale bar (ruler) or reference markers in the shots. For example, place a 1 m rod or checkerboard near the subject and use that known distance in postprocessing to scale the model. This isn’t always necessary, but for measurement applications where accurate dimensions are required, prepare these aids.

After assembling the equipment, select the target (site or object) and plan your on‑site photography. The next section covers photography tips for SfM.

Photography tips: overlap and multiple viewpoints are key

The success of SfM processing depends greatly on the quality of photography. Beginners should keep the following points in mind.

• Sufficient photo count and overlap: To get a high‑quality 3D model, take many photos that cover the entire subject, ensuring 60–80% or more overlap between consecutive images. If overlap is insufficient, the software may not find matching points between images and fail to reconstruct the model. It’s safer to err on the side of taking more photos than you think necessary.

• Photograph from various angles: Capturing the same subject from different angles and heights provides three‑dimensional information. Avoid photographing only straight above or straight on; instead, shoot from oblique angles that surround the subject. For large subjects like buildings or terrain, plan photo positions to capture the subject from all sides.

• Watch focus and blur: Ensure all photos are in focus. Blurry or out‑of‑focus images hinder analysis, so use a tripod, set a fast shutter speed, or use smartphone image stabilization. In bright outdoor conditions, shooting is easier, but in dim indoor or evening settings, add lighting rather than raising ISO too high (which increases noise).

• Keep exposure and settings consistent: If images vary greatly in brightness or color, the software may struggle to recognize the same subject. Try to keep camera settings (exposure, white balance) consistent so the image set is uniform. If using auto mode, correct for changes if settings shift during shooting.

• Get close to the subject: It’s better to get as close as possible rather than zooming from far away to capture fine details. For very large structures, include some wide shots to cover the overall shape along with closer detail shots.

Following these tips will produce image data that SfM software can process reliably. Once photography is complete, proceed to the SfM processing steps in software.

Workflow of 3D model generation with SfM software

Post‑photography processing is handled by software, but understanding the workflow helps deepen comprehension. A typical SfM processing flow is as follows.

• Import photos: Load the multiple photos into the SfM software. Camera parameters (lens focal length, sensor size, etc.) may be auto‑detected, but check settings as needed.

• Feature detection and matching: The software automatically detects feature points within each image (high‑contrast patterns, corners, etc.) that serve as alignment cues. It then matches the same points (correspondences) across multiple photos. The more common feature points found across image pairs, the higher the later computational accuracy.

• Camera pose estimation (photo alignment): Using feature matches, the software back‑calculates where and at what angles each photo was taken. This step estimates the camera’s external parameters (position and orientation) and computes the 3D coordinates of the detected points (a sparse point cloud). In short, this recreates camera viewpoints and a rough shape of the subject as a set of points.

• Dense point cloud generation: Based on the estimated camera poses and sparse point cloud, more detailed pixel‑level analysis produces a dense point cloud. This provides a detailed representation of the subject’s shape, capturing building details and terrain undulations.

• Create 3D mesh and texture: If needed, generate a polygon 3D mesh from the point cloud and apply texture mapping using the photos so the model looks realistic. Some software can also output an ortho‑mosaic image (nadir‑view photomap) for ground surfaces.

• Scale adjustment and coordinate alignment: Align the model to real‑world dimensions and coordinate systems. If a scale bar was included in the images, use that known distance to scale the model. For georeferencing in survey coordinates, measure a few ground points on site and assign their coordinates to corresponding model points—this aligns the model to absolute coordinates (called georeferencing).

• Export and use: Export the completed 3D model or point cloud in the required formats. For construction, import point clouds into CAD for volume calculations; for cultural heritage, publish 3D models online—there are many use cases.

Steps and terminology vary by software, but the overall flow is as above. Beginners may rely on automatic processing initially, and as they become familiar, adjust parameters in each step to achieve higher accuracy.

Tips for obtaining high‑precision models

To get satisfactory 3D models even on your first try, follow these accuracy improvement tips.

• Use high‑resolution, low‑distortion images: Image quality directly affects model accuracy. Shoot at the highest practical resolution and correct lens distortion in wide‑angle lenses via calibration to compensate for lens characteristics.

• Ensure sufficient overlap and multi‑view coverage: As noted earlier, higher overlap helps reliable feature matching. Combining images from many angles reduces blind spots and produces a precise model. The basic rule for SfM success is "take lots of overlapping shots from various angles."

• Choose subjects with appropriate features: SfM tracks surface features, so uniform surfaces like blank walls or glass are problematic. If the target lacks texture, add markers to create artificial reference points.

• Balance data volume and processing time: Too many photos can lead to long processing times or PC freezes. Start with a few dozen images, then increase as needed. Some software offers settings that temporarily downsample images for feature detection to reduce load—use these wisely.

• Consider partial retakes: If the result shows sparse or noisy point clouds in specific areas, it may mean insufficient photos for that region. If possible, retake concentrated shots of the problem areas and reprocess to improve quality.

These practices will raise the accuracy and completeness of SfM models.

SfM use cases: from everyday applications to full‑scale surveying

3D models from SfM processing are used across various fields. Here are examples from beginner‑friendly applications to professional surveying.

• Construction and civil engineering site surveying: Use drones or smartphones to photograph sites, create 3D models for as‑built management and volume calculations. SfM enables planar capture of site shapes at once, improving efficiency over measuring many individual points.

• Cultural heritage and archaeological documentation: Projects around the world photograph historic buildings and ruins for SfM processing to create detailed digital archives. For fragile cultural assets, non‑contact 3D capture from photos is highly valued.

• Accident scene analysis: SfM can reconstruct traffic accidents or disaster sites for post‑incident analysis and reporting. Situations difficult to convey in 2D photos are intuitively understood in 3D models, aiding investigation.

• Game and VR content creation: Scan real objects (sculptures, small items) with SfM to create 3D assets for games and VR. It’s an easier way to obtain realistic geometry than manual modeling, attracting interest in creative industries.

• Education and research: Use SfM for academic surveys of terrain or vegetation, or architectural students importing models of their scale models into design workflows—education and research are expanding SfM applications.

SfM’s application range is wide, from hobby projects for beginners to professional uses in business.

Precautions to avoid failure

Finally, here are precautions to help ensure smooth SfM success and reduce the risk of failure.

• Clear the area before shooting: Moving people or vehicles can change scenes between photos and disrupt feature matching. Shoot in as static an environment as possible and remove obstructing objects beforehand.

• Avoid mixing unrelated photos: Only use photos of the same subject in processing. Including unrelated images can prevent correct reconstruction. Organize photos into folders after shooting.

• Check intermediate results: Monitor alignment results (camera poses and sparse point cloud) as processing proceeds. If a camera position is clearly off, exclude that photo or consider retaking it. Early intervention reduces rework.

• Prepare for software crashes: Save progress if the software allows incremental saves. Long processes can fail, and losing an overnight run is a common frustration—save periodically.

• Observe laws and privacy: When flying drones, comply with aviation laws. Avoid photographing restricted or confidential sites, and be mindful of privacy when others may appear in images. Check regulations appropriate to the purpose and location in advance.

By keeping these points in mind during preparation, shooting, and processing, you can avoid major failures. Although the first attempts may require trial and error, practice will build skill.

Conclusion: make better use of photos with SfM processing

SfM processing, which creates 3D models from photos, is an attractive technology that beginners can implement by following the right steps. Even without special sensors, you can get high‑precision 3D data with familiar cameras and a PC, and applications will only expand. Start with a small subject and try SfM processing.

If you want to apply SfM‑created 3D models to real field measurements or business use, consider simplified surveying with LRTK. LRTK combines a smartphone with high‑precision GNSS to seamlessly perform photogrammetry and position measurement. It’s easy for beginners to use and can make SfM‑based 3D measurement more straightforward and reliable. Use the latest technologies to fully unlock the potential of your photos.