How to Use SfM in Cultural Heritage Surveys? 6 Photography Steps to Avoid Failure

In cultural heritage survey sites, recording methods are required that minimize damage to the object, accurately preserve its current condition, and make it easy to recheck later. SfM has been attracting attention for this purpose. This technique, which reconstructs three-dimensional shapes from photographs, is increasingly being used in a variety of cultural heritage surveys, such as recording stone monuments, buildings, archaeological remains, wall surfaces, sculptures, and excavated artifacts.

On the other hand, SfM is not a matter of "just taking lots of photos for the time being." If photographs are taken poorly, problems can occur such as distorted shapes, rippled surfaces, missing parts, and reduced reliability of dimensions. In cultural heritage documentation, merely reproducing a plausible appearance is insufficient; photography must be carried out with awareness of reproducibility, comparability, and preservation.

In this article, aimed at practitioners searching for information on "SfM cultural heritage", we organize from a practical perspective the key concepts to keep in mind when using SfM in cultural heritage surveys and six shooting procedures to avoid failure. We provide a comprehensive explanation covering preparations before shooting, on-site decision-making, common pitfalls that can easily degrade recording accuracy, and post-shooting checkpoints, so the content is useful both for those introducing SfM for the first time and for those already using it who want to stabilize quality.

Table of Contents

• Why SfM Is Used in Cultural Heritage Surveys

• Survey policies that should be decided before shooting

• Step 1 Set the scope and purpose of the target first

• Step 2: Adjust the lighting conditions to minimize shadow variations

• Step 3 Perform a full orbit capture with sufficient overlap to avoid gaps.

• Step 4 Vary height and angle to avoid missing parts of the three-dimensional shape

• Step 5 Increase recordability by being aware of scale and positional relationships

• Step 6 Conduct on-site verification and don't hesitate to retake photos

• Typical cases that tend to degrade SfM accuracy in cultural heritage surveys

• On-site operations to further enhance the recording value of SfM

• Summary

Why SfM Is Used in Cultural Heritage Surveys

SfM is an approach that, using common points visible in multiple photographs as clues, estimates the relative positions of the cameras and the shape of the object, and generates the data that serve as the basis for three-dimensional models, point clouds, and orthophotos. The reason SfM is attracting attention in the field of cultural heritage surveys is that it makes it easy to record both shape and surface condition together while minimizing contact with the object.

Cultural heritage has value not only in its form but also in information such as surface weathering, losses, cracks, incisions, seams, construction marks, and traces of repairs. Conventional two-dimensional photographs alone have sometimes failed to convey spatial relationships and depth adequately. Conversely, plans and sections alone can miss surface texture and localized deterioration. SfM is well suited as a method to fill the gap between photographic records and geometric records, and its strength is that it is easy to use as foundational material for post-survey analysis, report preparation, comparisons over time, and conservation planning.

Also, in cultural heritage surveys there are often locations that are difficult to enter repeatedly, as well as places that require scaffolding, access restrictions, or conservation-related precautions. In such environments, people frequently regret information they failed to capture on site. If you secure a sufficiently large set of photographs for SfM, the chances of being able to review the site three-dimensionally afterwards increase, reducing the burden of additional inspections. This is important not just as an efficiency measure, but in the sense of bringing back as richly as possible the conditions that can only be obtained on site.

However, SfM is not infallible. When imaging conditions are unstable, the fine reliefs and edges characteristic of cultural heritage can be flattened or, conversely, rendered as noisy artifacts. In particular, glossy surfaces, monotonous patterns, strong backlighting, deep shadowed recesses, thin protrusions, and continuous smooth planes are elements whose shapes are difficult to estimate reliably from photographs. That is why, to succeed with SfM in cultural heritage surveys, you must not rely solely on the software; you need the mindset of ensuring quality at the time of capture.

Investigation Policies to Determine Before Shooting

Failures in SfM often begin not at the time of shooting but from inadequate prior planning. In cultural heritage surveys, one must first clarify what the recording is for. Whether it is for a preservation ledger, for pre- and post-restoration comparison, for condition monitoring, for research material, or for visualization for public exhibition, both the required density of photographs and the necessary reproduction range will vary.

For example, if you only need to capture the overall shape, a photography plan that reliably covers the entire subject is important. On the other hand, if you want to see surface chips or traces of fine detail, relying solely on overview shots is not enough; you need a setup that incorporates close-up photography. Moreover, if you may later perform dimensional verification or positional comparisons, you must consider scale and its relationship to coordinates from the outset.

Another important matter is checking the object's preservation state and the on-site conditions. In places that must not be touched, places you cannot approach, places where long-term illumination should be avoided, or locations where passage and work flow are restricted, the freedom to photograph is greatly limited. Therefore, SfM in cultural heritage surveys should be regarded not simply as a photographing technique but as practical work that integrates preservation considerations, site management, and recording design.

When planning a shoot, it's risky to leave decisions about which directions to shoot from, at what heights, and in what order solely in your head. Walk around the subject and check for blind spots, tight spaces, the way shadows fall, background clutter, and the safety of your footing; if necessary, write the shooting order down on paper to reduce mistakes. In cultural heritage surveys the cost of re-shooting tends to be high, so five minutes of preparation before shooting can affect several hours of subsequent work.

Step 1: Define the scope and objectives first

The first step is to clearly determine the scope of what you will photograph. This may seem obvious, but in practice it is a stage that often begins ambiguously. If you do not decide in advance whether you want to document the entire subject, focus on specific design details or areas of deterioration, or also record its relationship with the surrounding environment, the planning of the set of photographs will become inconsistent.

One common mistake in cultural heritage surveys is that, on site, thinking “let’s just shoot broadly for now” leads to an insufficient photo density of the crucial subject. If you include too much of the surroundings, the subject itself appears small and its surface features become faint. Conversely, if you get too close, the positional relationships with the whole won’t connect, and it may only be possible to reconstruct them as separate chunks. What’s important is to separate the roles of overview, mid-range, and close-up shots, and plan so they ultimately connect as a single continuous set of photographs.

At this stage, rather than viewing the subject as a surface, it is better to decompose it and think of it as a three-dimensional solid. Check not only the front but also the sides, the top, recessed areas, protrusions, missing or damaged sections, and stepped areas to determine where the surfaces necessary for shape restoration are located. Cultural heritage objects often have complex contours, and there are frequently places that, even though you think they are visible, are unexpectedly not captured. In particular, areas under eaves, carved recesses, the insides of joints, and overlapping parts are prone to be missing, so it is important to identify the priority areas at the outset.

Also, it is necessary to work backward from how the deliverables will be used. If the primary purpose is still images to be included in a report, shooting with visually pleasing angles can be meaningful. However, if there is a possibility that comparisons of deterioration or dimensional checks will be carried out later, geometric continuity should be prioritized over aesthetics. In SfM for cultural heritage surveys, ensuring the continuity required for documentation comes first, rather than photographing with the sensibility of artwork photography.

Step 2 Adjust the lighting conditions to reduce uneven or harsh shadows

Lighting has a major influence on the quality of SfM. In cultural heritage surveys, the surfaces of the subject often have fine bumps and variations in texture, so moderate shading helps with the extraction of feature points. However, if shadows are too strong or the lighting changes significantly during capture, the same area can appear as a different object, and consistency can easily break down.

Particular attention should be paid to times when direct sunlight is strong. Under strong sunlight, the contrast between bright and dark surfaces becomes extreme, increasing areas that are blown out (overexposed) and areas that are crushed into black (underexposed). Traces and fine boundaries carved into the surfaces of cultural properties are easily lost due to this blowing and crushing. Furthermore, if the photographer’s own shadow or the shadows of surrounding structures shift position between photos, the appearance of the same spot changes and 3D reconstruction becomes unstable.

Ideally, you should shoot everything together under conditions where the light is stable. Soft light on cloudy days tends not to produce strong shadows and often works well with SfM in cultural heritage surveys. Even indoors, be careful when strong sunlight is streaming in from a window or when there are localized lighting unevennesses. If one side is extremely bright, the amount of information will be biased toward only part of the surface. If the weather suddenly changes during shooting, or lights are turned on or off, the consistency of the photo set will decrease. It is important to be mindful of completing a full circuit of shots under the same conditions as much as possible.

What's important here is that making the texture of cultural heritage look good and creating conditions suitable for SfM are not necessarily the same. Visually, stronger shading can make forms appear more three-dimensional, but for SfM, extreme highlights and deep shadows can get in the way. For documentation photography, prioritizing a consistent appearance between images rather than aiming for striking photos will produce more stable results.

Glossy or damp surfaces also tend to change their reflections depending on angle, which can make the same pattern appear to be in a different location. Wet stone, metallic reflections, and the sheen of protective coatings can easily become sources of error, so it is important not to neglect observing the shooting conditions. In cultural property surveys it is often difficult to treat the object, so the photographer’s ability to read the light can make a difference in quality.

Step 3: Perform loop photography with attention to overlap to ensure complete coverage

The most basic and important aspect of SfM photography is the overlap between photos. If adjacent photos do not share a sufficiently large common area, the connections between images become weak, causing the model to break apart or the spatial relationships to become distorted. In cultural heritage surveys, failures of the type “only part of the subject was photographed” occur less frequently than failures where “even though many photos were taken, they don’t connect.” Much of the cause of the latter is insufficient overlap.

For circular (orbit) shooting, the basic principle is to move gradually around the subject and take consecutive photos at a similar scale. If your steps are too large or you change direction abruptly, the overlap between consecutive photos decreases and the alignment becomes unstable. Taking a few shots in front of a cultural property, then jumping to the side and then to the back is riskier than it appears. SfM prefers smooth trajectories to stepping-stone–style shooting.

What you should focus on here is not the sheer number of photos but creating a dense, highly continuous sequence of images. Increasing the number of nearly identical photos excessively does not improve efficiency, and if the necessary overlap is lacking, the results will not be stable no matter how high the resolution. In cultural heritage surveys there is also pressure to shorten shooting time, but cutting back on the minimum required continuity tends to lead to revisit or re-shooting.

Also, care must be taken with the background. If there are walls, scaffolding, or grid-like structures around the subject with repeating patterns similar to the subject, features can be pulled toward the background. For this reason, it is important not to let the subject’s position shift greatly within the frame and to keep the main subject captured steadily. It is not necessary to remove the background completely, but if the proportion of the subject in each photo changes drastically, alignment can easily be disrupted.

When the shape of a cultural property is large, stability is improved by shooting with layers in mind: not just one circuit, but several circuits at the same height, and then several more at slightly different heights. This is especially true for subjects with significant vertical information—such as buildings, stone walls, small shrines, stone monuments, and sculpture pedestals—where using only a single height tends to weaken the connections between upper and lower parts. Although shooting in circuits may look simple, in practice which layers you connect and how you connect them determines the final quality.

Step 4 Vary height and angle so you don't miss three-dimensional shapes

When documenting cultural heritage with SfM, collecting only frontal photographs can sometimes fail to produce sufficient 3D reconstruction. In particular, recesses, protrusions, constrictions, deep carvings, and sloping surfaces tend to lack shape information unless there is variation in viewpoints. Therefore, in Step 4 it is important to deliberately vary the height and angle.

For example, with stone monuments, a set of photos taken all around at eye level tends to lack information about the top surface and often makes the boundary with the pedestal ambiguous. Conversely, shooting only from higher positions weakens the continuity of the sides. What is needed in cultural heritage surveys is not beautiful photos from a single viewpoint but a set of images that support the subject’s three-dimensional structure from multiple directions. Combining a slightly downward angle, an angle close to horizontal, and, when necessary, a slightly upward angle increases the stability of the reconstruction.

When changing angles, be careful not to change the shooting conditions too abruptly. It's best to change them gradually while maintaining continuity—for example, keep the camera level and centered on one pass and slightly bird's-eye on the next. If you suddenly mix in shots from an extreme high oblique angle, they may be treated as a separate group if the overall continuity is weak. Changing height and angle is important, but you must be aware that those changes also rely on the continuity with the photos before and after.

Cultural heritage often exhibits shape changes that are easy to miss with the naked eye. Shallow undulations in carvings, chipped edges of stone, steps at joints, surface bulges, and minute tilts can appear or disappear depending on the camera angle. To record these consistently, the same spot must be captured from different angles. In other words, increasing the number of photographs is not simple redundancy but a way to reduce bias in how features are perceived.

Also, in confined sites you may not be able to stand back far enough. In such cases, rather than forcing a wide shot, it is more effective to divide the subject into multiple bands and photograph them sequentially while keeping in mind a composition that will be stitched together later. If the upper, middle, and lower bands overlap well, the chances of successful geometry reconstruction increase even in a limited space. In SfM for cultural heritage surveys, flexibility is required to design variations in viewpoints to suit site conditions.

Step 5 Enhance documentation by being aware of scale and spatial relationships

SfM can reconstruct three-dimensional shapes from photographs, but how much value the model has as documentation depends on the approach to scale and positional relationships. Even if you produce a visually pleasing three-dimensional model, if the reference for size is ambiguous and it cannot be correlated with other records, its usefulness as survey documentation is limited.

In cultural property surveys, it may be necessary to compare different periods, to check against plans and cross-sections, and to confirm changes before and after restoration. Therefore, from the moment of photographing, it is important to decide what to use as a reference and over what area to maintain a consistent positional relationship. The term "scale" here is not just for knowing lengths, but also serves as the basis for overlaying other records later.

On site, attention tends to be focused only on the object itself, but in practice it is important to manage stable reference points in the object's surroundings. Thinking in advance about which positions are easy to reconfirm, which directions correspond easily with site drawings, and which viewpoints can be readily reproduced during future re-measurements will greatly increase the value of the records. Because cultural properties are materials handled over the long term, it is necessary to design processes that are not completed by on-site treatment alone.

The relationship between partial photography and overall photography is also important. For example, even if you densely photograph decorative or damaged areas, they become difficult to use if it is not clear later where they belong in the overall object. Properly connecting the group of photos that capture the whole with the group that documents the details is fundamental to SfM operations in cultural heritage surveys. Even if the details are beautifully reconstructed, if their relationship to the whole is lost, the persuasive power of the record is weakened.

In cultural heritage surveys, it is also effective to associate data with on-site location information. When dealing with multiple subjects in a single survey or later examining their relationships with surrounding features, it is helpful not to rely solely on models derived from photographs but to think in terms of linking them to site coordinates and control point information. This makes it easier to move beyond mere three-dimensional visualization and develop records that can be used to understand the site as a whole.

Step 6: Conduct on-site checks and don't hesitate to retake photos

What’s crucial at the end of SfM shooting is to always perform an on-site check before leaving the site. In cultural heritage surveys, even if you later notice that “part of it was missing,” “required surfaces did not connect,” or “dark areas were lost,” you may not be able to revisit under the same conditions. Therefore, shooting does not end when the shutter is released; it should be considered a single process that includes verification.

In on-site checks, the first thing to verify is whether there are continuous photos sufficient to go all the way around the entire subject. Next, confirm there are no missed shots of the top, bottom, recessed areas, areas of deterioration, or boundary areas. Also check for any bias, such as a particular group of images being extremely dark, overly bright, blurred, or taken from too great a distance. Even if the number of photos is adequate, the whole set can become unstable if continuity is broken somewhere, so it’s important to review from the perspective of continuity.

In cultural heritage surveys, when you're on site you tend to want to decide "this much photography should be enough." However, that intuition is surprisingly unreliable. Information you can mentally fill in when viewing the real object often isn't present once you bring back a set of photographs. In particular, the simpler a surface's shape appears, the easier it is to become complacent, and during post-processing you may lack the clues needed and the result can fail. Smooth stone surfaces, uniformly colored walls, and areas with repetitive patterns are safer to document by deliberately ensuring greater coverage.

It's also important not to hesitate to retake shots. On site, constraints of time, physical strength, and weather can make it feel troublesome to shoot again after packing up. However, in SfM for cultural heritage surveys, a few additional minutes of shooting can determine the success or failure of the entire subsequent process. Capture any areas you are unsure about, and cover uncertain parts before conditions change—those judgments lead to practical, professional-quality results.

On-site verification should not be done by merely looking at images; it is better to approach it from the perspective of "can a third party understand the subject later from this set of photographs?" Don't be satisfied by assuming you already know the site; being mindful of whether the materials are sufficient for another person in charge or for your future self to reuse will improve the completeness of cultural property documentation.

Typical examples that tend to reduce SfM accuracy in cultural heritage surveys

When operating SfM in cultural heritage surveys, failures tend to concentrate in specific situations. One common issue is excessive variation in camera-to-subject distance. If one photo is taken close to the subject and the next is shot from much farther away, the subject’s size will vary greatly between images, making feature matching unstable. Mixing overview shots and close-up shots is not a problem in itself, but without intermediate-distance photos to bridge them, the model can easily become fragmented.

Another common issue is underestimating monotonous surfaces. Stone walls, plastered surfaces, worn surfaces, and other areas with little visual information make it difficult to determine location from photos. For such surfaces, you need to photograph them continuously while including distinctive surrounding elements. If you only photograph areas with few features, the connections will be weak.

Furthermore, placing too much emphasis on details can cause their relationship to the whole to be lost. Damaged areas and patterned sections of cultural heritage are important, but even if you densely photograph only those parts, the material becomes difficult to use for research unless there is a bridge to the overall view. SfM is often praised for its ability to reproduce fine detail, but in cultural heritage surveys it is important to balance both the details and the whole.

Be careful when there are many moving elements around the subject. Passing people, swaying plants, shifting shadows, and fluttering cloth or protective coverings can change between photos and introduce noise. Even if the cultural property itself remains still, unstable surrounding conditions will degrade reconstruction quality. It is necessary to be conscious of managing the entire shooting environment, not just the subject.

Finally, another common mistake is not shooting with the processing workflow in mind. If you mix in tilted photos, extremely dark photos, or photos containing many unwanted objects with the idea that “we’ll fix it later,” the effort required for selection and reconstruction increases. In cultural heritage surveys, SfM is a method that becomes easier the more you prepare on site. The care taken at the time of shooting directly translates into the stability of the results.

Field operations to further enhance the documentation value of SfM

The significance of using SfM in cultural heritage surveys is not the mere creation of a three-dimensional model. The true value lies in leaving a record that can be referenced over time and in linking and utilizing it with other materials and coordinate information. For that reason, it is important not to treat the process as complete with photography alone, but to design the entire on-site workflow.

First, it is important to organize and record basic information such as the photography date, weather, coverage area, subject name, camera direction, and areas of focus. In cultural property surveys, data is often reviewed later by different personnel, and the intent behind photographs can be lost when only the photos are provided. In particular, simply knowing which areas were prioritized for photography and which parts were the subject of damage or deterioration inspections can greatly affect the reusability of the data.

Also, if you anticipate comparing multiple time points, it is effective to set up procedures that make it easy to re-photograph from the same relative positions. Because cultural properties are managed with long-term preservation in mind, photography should not be a one-off activity; you need to adopt a recording system that links to subsequent surveys. When the structure of the photo sets and the organization of site coordinates are in place, it becomes easier to track changes over time and to compare pre- and post-restoration.

A useful idea here is to combine SfM photographic records with high-precision positional information. For example, by accurately recording on site the surrounding reference points and the locations of the cultural property under investigation, three-dimensional records derived from photographs become easier to handle as practical documentation. When you want to organize information that includes conservation management and relationships with the surrounding environment, the accuracy of positional information carries significant value.

In situations where you want to streamline on-site coordinate verification and recording tasks, operations that combine a smartphone-mounted high-precision GNSS positioning device like LRTK are effective. SfM excels at reconstructing shape from photographs, but if there is a mechanism to smoothly establish reference positions on site and the locations of survey targets, the potential uses of the records expand further. In cultural heritage surveys, not only is recording the object itself important, but it is also essential to record where, over what extent, and under what conditions the recording was made. By leveraging LRTK, you can more easily improve the efficiency of control-point surveying and on-site coordinate verification, making it easier to integrate SfM records with site management.

Summary

When using SfM for cultural heritage surveys, it is crucial to clarify what you want to preserve and at what quality before considering equipment or processing procedures. Then fix the target area, adjust the lighting conditions, perform circumferential shooting with attention to overlap, supplement three-dimensional information by varying height and angle, and thoroughly verify things on site while being mindful of scale and spatial relationships. By following these six steps, you can obtain SfM data that are useful not only for visual purposes but also as records.

Each cultural heritage item has different conditions, and the same approach does not necessarily apply every time. However, many failures can be traced to common causes: lack of continuity during shooting, inconsistent lighting conditions, disconnection between overall views and details, and insufficient on-site verification. For that reason, in SfM for cultural heritage surveys, pre-shoot planning and on-site judgment determine the results.

If you want to make 3D documentation of cultural heritage more practical, it is important to develop not only photograph-derived shape records but also on-site positional data and verification of reference points. By combining high-precision positioning devices such as LRTK with SfM-based records as a foundation for linking documentation to investigation, conservation, comparison, and sharing, it becomes easier to determine site coordinates and streamline basic surveying. At sites aiming to raise the quality of cultural heritage survey records, treating SfM and positioning operations as an integrated workflow rather than separating them will lead to future practice.