Field Photography Protocol for Heritage Photogrammetry

Why Capture Decides Everything

Structure-from-Motion works by finding the same physical point across many overlapping photographs and triangulating its 3D position from the differing camera viewpoints. This mechanism sets three non-negotiable capture requirements: every point on the surface must appear sharply in multiple images (so it can be matched), the images must overlap enough that matching is reliable, and the surface must carry enough visual texture for distinct points to exist at all.

When any of these fails, the failure is silent at capture time and expensive at processing time. A slightly soft image looks acceptable on the camera's rear screen but produces weak feature matches. A gap in overlap looks like adequate coverage to the eye but leaves a hole the algorithm cannot bridge. A uniformly-lit white marble surface looks perfectly photographed but has no texture for the algorithm to lock onto.

Camera Settings for Heritage Photogrammetry

The camera should be in full manual mode. Automatic exposure, autofocus drift and auto white balance all introduce inconsistency between frames that degrades feature matching and colour fidelity. The goal is that every image of a given surface is captured with identical settings, so the only thing that changes between frames is the viewpoint.

The Exposure Triangle for Sharp Images

The three exposure controls must be balanced in a fixed priority order for photogrammetry, which is different from the priority a general photographer would use. First, set aperture to f/8–f/11 for depth of field — this is fixed and non-negotiable for a sharp surface. Second, set the shutter speed fast enough to eliminate motion blur — for handheld capture this means 1/125s or faster, faster still with longer lenses. Third, and only third, raise ISO as the last resort to reach a correct exposure once aperture and shutter are set.

This ordering means that in low light — a shaded courtyard, an overcast evening, a partly-enclosed mandapam — you will often need a tripod. A tripod removes the shutter-speed constraint entirely (the camera is static, so a slow shutter causes no blur), letting you keep low ISO and a small aperture. For any interior or low-light heritage capture, a tripod is not optional equipment; it is the difference between a usable and an unusable dataset.

Overlap and Coverage Rules

Overlap is the proportion of a scene shared between two adjacent photographs. Too little overlap and the algorithm cannot reliably link the images; the model breaks into disconnected fragments or fails to align. For heritage work the safe minimums are higher than for flat aerial mapping because heritage surfaces are geometrically complex and self-shadowing.

• Flat facades and walls: 70% minimum overlap between adjacent images, both horizontally within a strip and vertically between strips. Overlap between strips is as important as overlap within a strip.
• Carved ornament and relief: 80% or more overlap. Deep recesses are seen from fewer angles, so the surrounding coverage must be dense enough that every recessed point still appears in several images.
• Freestanding sculpture (photographed in the round): 80%+ overlap with the camera moved in a smooth arc; each image should overlap substantially with the one before and after it around the full circuit.
• Corners and edges: shoot additional convergent (angled) images at every corner where two surfaces meet. Corners captured only with square-on images are a frequent source of geometric error.
• Never zoom or crop to change framing between frames — move your feet to keep the subject at a consistent size in the frame; changing framing by zooming breaks the geometric consistency the algorithm relies on.

Lighting: The Silent Failure Point

Lighting is the capture variable most often ignored and most damaging when wrong. The ideal light for photogrammetry is flat and diffuse — soft, even illumination with no hard shadows and no bright highlights. This is the opposite of what makes an attractive architectural photograph, and it is a discipline that architectural photographers often have to unlearn for documentation work.

Hard direct sunlight is the enemy. It casts deep shadows into carved recesses (hiding the very detail heritage documentation exists to record), blows out highlights on light-coloured stone (destroying texture), and — critically — the shadows move as the sun moves, so a surface photographed over an hour has inconsistent shadow positions between frames, which confuses feature matching.

Capture Protocols by Subject Type

Different heritage subjects demand different capture geometry. The following protocols are the field-tested starting points for the four subject types most common in South Indian heritage documentation.

Scale, Reference and Colour Control

A photogrammetric model built from photographs alone has no inherent scale — the algorithm reconstructs geometry and camera positions but cannot know the real size of the subject. Scale must be introduced through either ground control points measured by survey instruments, or through calibrated scale bars placed in the scene and visible in the images. For measured heritage documentation, ground control measured with a total station is the rigorous method; for smaller objects, a photogrammetric scale bar with coded targets is efficient and accurate.

Colour and tonal accuracy matter for heritage records because condition interpretation — distinguishing staining, biological growth, salt efflorescence and material change — depends on faithful colour. Include a colour reference target (a grey card or a colour checker chart) in at least one frame of each lighting setup, and set white balance to it. This allows accurate, repeatable colour correction of the whole dataset in RAW processing before the model is built.

• Place at least two scale bars or three control targets in the scene, well distributed, so scale can be set and independently checked.
• Include a colour checker or grey card in a reference frame for each distinct lighting condition; note which frame it is in.
• Photograph a written or printed slate at the start of each subject: site name, element, date, operator — this labels the dataset unambiguously in the field.
• For inscriptions and worn carving, capture a separate raking-light (low angle) photographic pass for legibility; keep it separate from the flat-lit model images, which need even light.

In-Field Quality Assurance Checklist

The purpose of in-field QA is to catch a fatal problem while you can still fix it — before leaving the site. Run this check before you consider a subject complete and pack up.

1. 1Sharpness: review a sample of images at 100% zoom on the rear screen; carved edges and stone grain must be crisp.
2. 2Exposure: check the histogram — no large clipped highlight or shadow spikes; the surface should sit in the mid-tones.
3. 3Coverage: mentally (or on a sketch) confirm every face, recess and corner of the subject has been photographed from multiple angles with overlap.
4. 4Overlap: confirm you moved in small enough steps that adjacent frames share most of their content.
5. 5Convergent images: confirm you shot oblique/angled images across the surface, not only square-on.
6. 6Scale and reference: confirm scale bars/control targets and a colour reference are captured and visible.
7. 7Consistency: confirm settings and focus did not change mid-sequence (a common error after a battery or lens change).
8. 8Redundancy: when a subject is important and access is difficult, capture a second full pass — storage is cheap, re-mobilisation is not.

Common Capture Mistakes

• Shooting in JPEG to save space — JPEG discards tonal and colour data needed for consistent processing and condition interpretation; always shoot RAW.
• Trusting the rear screen fit-to-view — softness and blur are invisible until you zoom to 100%; the small preview hides the most common fatal problem.
• Photographing in bright direct sun — hard moving shadows and blown highlights are the leading cause of poor exterior models; wait for cloud or work in the softest light available.
• Using autofocus and letting it refocus between frames — the shifting focal plane changes image geometry; focus manually and lock it for the sequence.
• Only shooting square-on — square-on-only capture produces systematic doming distortion; always add a convergent oblique pass.
• Changing framing by zooming instead of moving position — this breaks geometric consistency; keep focal length fixed and change viewpoint by walking.
• Forgetting scale reference — a beautifully detailed model with no scale bar or control cannot be used for measured documentation; it is a picture, not a survey.
• Attempting dark interiors handheld without added light — SfM needs texture and light; a dark handheld sequence produces blur and failed matches. Use a tripod and controlled lighting, or use laser scanning.

Professional Practice

In professional heritage documentation, capture discipline is what separates a repeatable service from an unpredictable one. The practitioner who captures slowly and checks at 100% in the field delivers reliably; the one who shoots quickly and hopes discovers failures back in the office, when the fix is a costly return visit or an apology to the client.

Capture protocol should be written into the method statement for each project and matched to the required output accuracy. A model destined for a 1:50 measured elevation demands stricter overlap, control and lighting than a model captured for general visual record. Agreeing the intended accuracy before the site visit determines the capture protocol — and prevents the situation where images captured for a visual record are later expected to deliver a measured drawing they cannot support.

For sites that will be documented repeatedly over years — condition monitoring of a deteriorating monument, for example — recording the exact capture setup (positions, settings, control layout) allows future surveys to be captured comparably, so that change can be measured reliably between visits rather than confounded by inconsistent capture.