3D Modelling for Heritage Documentation: Methods, Software and Outputs

Types of 3D Model in Heritage Documentation

Heritage 3D models fall into three categories defined by their purpose, their data source and their geometric nature. Each serves different documentation functions and is produced by different software workflows.

Survey-derived models (photogrammetric meshes, point clouds, scan-to-surface models) are the closest to raw measurement data. They represent the building's actual surface geometry as captured by photogrammetry or laser scanning, without interpretation or modelling decisions. They are dense, detailed and accurate to the as-found condition. They cannot easily be queried for building elements ('what is the floor area of the mandapam?') or updated when conditions change, but they carry the full resolution of the original survey.

Parametric HBIM models are constructed by modelling building elements (walls, columns, arches, floors, roofs) over survey data as reference. Each element is an intelligent object with properties (material, dimensions, layer, historical period, condition) that can be queried and updated. HBIM is the model type suited to conservation management, repair scheduling and heritage significance assessment. It involves interpretation — a modeller must decide how to represent a complex carved capital as a parametric element — and is slower and more expensive to produce than a survey mesh, but far more manageable for ongoing heritage management.

Visualisation models are purpose-built for rendering and public presentation. They may be based on survey data but are simplified for rendering performance, sometimes augmented with speculative reconstruction of missing or damaged elements, and prioritise visual quality over measurement accuracy. They are produced in Blender, SketchUp, 3ds Max or similar tools and rendered in Cycles, V-Ray or real-time engines like Unreal or Lumion.

Photogrammetric and Scan-Derived Mesh Models

The photogrammetric mesh is the most direct form of heritage 3D model: it is a triangulated surface representation derived computationally from overlapping photographs (Structure from Motion) or from a point cloud (Poisson or Delaunay meshing from TLS data). It represents the actual as-found surface geometry of the building at the time of survey, without modelling decisions, at a resolution limited only by the survey's point density.

The typical workflow for a heritage photogrammetric mesh: photographs are processed in Metashape or RealityCapture to produce a dense point cloud; the point cloud is meshed (generating a surface of triangular faces); the original photographs are projected onto the mesh to produce photorealistic texture (the UV mapping step); the textured mesh is exported in OBJ, FBX or GLTF format for delivery or further use. The result is a coloured, textured model that looks photorealistic and carries the actual surface appearance of the building.

For documentation purposes, photogrammetric meshes have important limitations. They represent the surface but not the structure: there is no distinction in the mesh between a wall, a floor, a carved decoration and the vegetation growing on the wall face. The mesh contains no semantic information. Structural analysis, material condition assessment and heritage significance mapping all require additional interpretation layers applied over the mesh — typically in a GIS or in specialist software such as Matterport or ArcGIS Indoors.

File sizes for full-detail heritage meshes are large: a complete exterior photogrammetric model of a medium temple complex may contain 100–500 million triangles and occupy 20–100 GB on disk at full resolution. Delivering this to a client or heritage agency requires either: (a) mesh decimation to a manageable resolution, with documented quality compromise; (b) a streaming viewer (Potree, Cesium ion, Sketchfab) that serves the full-resolution model progressively to a browser; or (c) a purpose-reduced version (10–20 million triangles) packaged with a documented note on the resolution trade-off.

HBIM Parametric Models

Heritage Building Information Modelling (HBIM) extends the BIM methodology — parametric modelling of building elements as intelligent objects — to historic structures. The key difference from new-build BIM is that in HBIM, the model is built to record what exists rather than to specify what will be constructed. Elements are modelled from survey data (photogrammetric mesh, point cloud, or total station measurements used as reference geometry), and properties are assigned to reflect their historical context, material composition, condition and significance.

The primary platform for HBIM in India and internationally is Autodesk Revit, though ArchiCAD (with the ARCHICAD Heritage Tools plugin) and Vectorworks Architect are also used. In Revit, a point cloud or photogrammetric model is inserted as a background reference layer; the operator then models walls, floors, columns, roofs and decorative elements by tracing over this reference, adjusting element dimensions to match the survey geometry. The result is a model in which every element is a Revit family object with dimensions, material and other properties.

HBIM models are the appropriate deliverable for heritage projects requiring: ongoing conservation management (the model can be updated as condition changes); schedule of condition or repair (element properties carry condition ratings and repair urgency); significance assessment (each element can carry a heritage significance designation); BIM coordination with structural engineers, MEP consultants and contractors; and formal handover to a heritage agency for long-term management.

Visualisation and Interpretation Models

Visualisation models serve public interpretation, stakeholder engagement, grant applications and heritage tourism. They may depict a building as it exists today, as it appeared at a specific historical moment (including reconstructed missing fabric), or as it might appear after a proposed conservation intervention. The priority is visual quality and communicative clarity rather than measurement accuracy.

Blender, 3ds Max and Cinema 4D are the primary tools for high-quality heritage visualisation modelling. Photogrammetric meshes can be imported directly and used as the basis for cleaned, optimised models with professional materials and lighting. For reconstructive work — where missing architectural elements are reinterpreted from historical sources or analogous examples — the modeller must distinguish clearly in any publication between recorded fabric and speculative reconstruction.

Real-time visualisation is an emerging application for heritage: Unreal Engine 5 (with its Nanite and Lumen systems for high-fidelity real-time rendering) and Unity are being used to create heritage virtual reality (VR) and augmented reality (AR) experiences. For South Indian temples, VR experiences allow viewers to understand the spatial progression of a temple complex — from entrance gopuram through prakaram to inner shrine — in a way that photographs and drawings cannot convey. Production of this kind of experience is a design and storytelling project as much as a technical one, requiring a different skill set from survey-grade 3D modelling.

Level of Detail (LOD) in Heritage 3D Modelling

Level of Detail (LOD) in heritage BIM, based on the BIM Forum's Level of Development specification (and similar to but distinct from RIBA stages), describes how geometrically complete and information-rich a modelled element is. Heritage projects typically reference a modified LOD scale that accounts for the difference between 'designed' and 'as-found' elements.

LOD 100 in heritage HBIM is a massing or conceptual representation: a block representing a building or part of a building with approximate dimensions but no detail. LOD 200 is a generic approximation with overall dimensions. LOD 300 is a specific representation of the element as surveyed: wall thickness and height from measurements, opening dimensions from survey. LOD 350 is LOD 300 plus interface and connection information. LOD 400 adds material specification and condition information. LOD 500 is the as-built (or as-surveyed) element with all construction and material information for facility management purposes.

For most conservation-grade heritage HBIM projects, LOD 300–350 for primary structural elements and LOD 200–300 for decorative elements is the appropriate target. Achieving LOD 400–500 across all elements of a significant historic building requires enormous modelling effort and is rarely justified by the available budget or the subsequent use of the model. Define the required LOD per element type in the project brief before modelling begins.

Software Overview

The choice of software depends on the type of model being produced and the downstream use case. No single platform serves all three heritage 3D model types well; professional projects typically use two or three tools in sequence.

Blender for Heritage Work

Blender is a free, open-source 3D creation platform used in heritage for mesh processing, photogrammetric mesh optimisation, visualisation rendering and heritage interpretation model production. Its node-based materials system allows photogrammetric textures from photogrammetry to be applied, adjusted and augmented with additional layers (weathering, overlay annotation).

For heritage mesh work, Blender's mesh editing tools allow manual repair of photogrammetric mesh artefacts (fill holes, remove isolated components, correct inverted normals), and its Decimate modifier provides controllable mesh reduction. The Geometry Nodes system (introduced in Blender 3.x) enables procedural generation of repeating architectural elements — useful for reconstructing missing sections of decorative cornice, colonnade or perimeter wall from survey data on an intact portion.

Blender's Cycles and EEVEE rendering engines produce high-quality photorealistic renders for heritage reports, publications and grant applications. For Tamil Nadu temple heritage specifically, Blender has been used by several institutions and independent researchers to produce rendered documentation models of Dravidian temple towers and sculpture, combining photogrammetric texture with additional hand-painted detail for missing or worn surfaces.

Revit and the HBIM Workflow

Autodesk Revit is the dominant BIM platform for professional HBIM work. The HBIM workflow in Revit begins with inserting a georeferenced point cloud (via Autodesk ReCap) or a photogrammetric mesh as a reference layer. The operator then models heritage elements by creating Revit Families (parametric element templates) matched to the types found in the building — for Indian temple architecture, this requires custom Families for elements such as gopuram stages, mandapam pillars, corbelled brackets and sculptural niches that are not in the standard Revit Family library.

The modelled elements are then given parameter values reflecting their heritage attributes: material (granite, sandstone, brick, lime plaster), estimated construction period, condition rating (good, fair, poor, critical), significance rating (high, medium, low) and recommended conservation action. These parameters can be exported as a Schedule (a tabulated list of all elements with their properties) for use in conservation management planning or grant applications to INTACH and NMA.

For Tamil Nadu temple work, the key challenge in Revit HBIM is the absence of native parametric families for Dravidian architectural elements. Producing accurate HBIM models of complex South Indian temples requires either creating custom Revit Families (a significant investment of time) or adapting imported 3D geometry as 'in-place families' — a less flexible but faster approach when standard parametric families are not available.

SketchUp for Heritage Interpretation

SketchUp occupies a middle ground between Revit's data-rich parametric modelling and Blender's polygon-centric approach. It is widely used in heritage for schematic massing models, presentation models for client and committee review, and site context models. Its learning curve is low, and models can be produced quickly from basic measurements.

For heritage documentation, SketchUp is most appropriate for: (a) schematic site models showing the spatial relationship of buildings within a complex; (b) presentation models for non-technical stakeholders (temple committee members, heritage committee, funding bodies); (c) early-stage design studies for conservation interventions; (d) AR/VR experiences targeted at general audiences using SketchUp Viewer or integration with platforms like XR platforms. SketchUp is not appropriate for conservation-grade measured drawing production or for HBIM with meaningful property data.

Photogrammetric Texture and Materials

One of the distinguishing advantages of photogrammetric 3D models over laser scan point clouds is colour: because the dense point cloud is derived from photographs, every point in the cloud carries the colour of the surface as photographed. When the point cloud is meshed, the photographs are projected onto the mesh surface to produce a photorealistic texture — the UV texture baked from reality rather than painted by a artist.

The quality of photogrammetric texture depends on: (a) the quality and consistency of the photography (even lighting, sufficient overlap, correct exposure); (b) the resolution of the source photographs (higher megapixel counts produce finer texture at a given model scale); (c) the quality of the mesh (good geometry makes texture mapping more faithful). For South Indian temple surfaces — carved granite, painted stucco, gilded elements — photogrammetric texture captures the actual patina, toolmarks and surface condition that an artist-painted texture could only approximate.

A common technique for high-detail heritage models is to use a lower-density mesh for the overall building geometry (for model management reasons) and then apply high-resolution photogrammetric texture patches to areas of particular detail — carved panels, inscription surfaces, decorative friezes. This approach, called multi-resolution texturing, concentrates rendering quality where it matters most while keeping overall model file sizes manageable.

Deliverable Formats for Heritage 3D Models

Heritage 3D models are delivered in formats appropriate to their purpose and the receiving software. The most common delivery formats and their appropriate uses:

Workflow: From Raw Data to Heritage Model

The standard workflow for producing a conservation-grade heritage 3D model from photogrammetry and TLS data: (1) Data capture — drone and ground photogrammetry for exterior, TLS for interior measured survey, total station for control network; (2) Point cloud processing — register all scans to a unified coordinate system, filter and clean, export to E57; (3) Photogrammetric processing — process drone and ground photographs in Metashape or RealityCapture to produce dense point cloud and textured mesh; (4) Model production — for mesh deliverable, decimate and repair the photogrammetric mesh in MeshLab or Blender; for HBIM, import point cloud as reference in Revit and model elements over it; (5) Texture application — bake photogrammetric texture onto the HBIM mesh surfaces for presentation-quality output; (6) QC — verify model dimensions against known total station measurements; (7) Export and delivery — export to agreed formats, prepare accuracy report.

A full HBIM modelling project for a significant temple complex — covering exterior gopuram, mandapam, main shrine, inner compound walls — requires 4–8 weeks of modelling time after survey data is received, depending on complexity, target LOD and the availability of appropriate Revit Families. For projects where only a textured mesh model is required (no parametric HBIM), production is faster: 1–2 weeks of processing and mesh preparation from survey data.

3D Modelling of South Indian Temples: Specific Considerations

South Indian Dravidian temple architecture presents specific 3D modelling challenges that do not arise in European or vernacular heritage contexts. The gopuram (gateway tower) is a receding pyramidal form with hundreds of stucco or stone sculptural figures arranged in register on all four faces. Photogrammetric capture must account for: the extreme vertical dimension (major gopurams reach 30–50 metres); deep shadows in the sculptural niches; variable surface — polished granite base, painted plaster above; and the difficulty of capturing the underside of projecting cornices from a nadir drone.

Supplementing drone nadir capture with oblique drone photography (camera angled at 45–60° and flown around the tower faces) is essential for gopuram documentation — nadir imagery captures the top surfaces of each stage but misses the vertical faces of the sculpture registers. Ground-based close-range photogrammetry for the lower 6–8 metres (accessible from ground level) combined with drone oblique imagery for the upper sections is the standard multi-method approach for major gopuram documentation.

For interior mandapam spaces — the colonnaded halls with carved granite columns — TLS is the primary capture method for measured drawing production. Photogrammetry works in interiors only when lighting is even (daylight or portable studio lighting) and ceiling height allows sufficient camera-to-surface distance. Narrow, dimly lit pradakshina paths (circumambulation corridors) around the main shrine are among the most difficult spaces to document: often too narrow for TLS setup, too dark and low for photogrammetry, and too confined for any drone operation.

The documentation of sculptural iconography — individual deities, narrative panels, decorative motifs — at high resolution requires close-range photogrammetry with controlled lighting, typically a DSLR camera on a monopod with a portable LED panel. This is treated as a separate documentation task from the architectural survey, producing high-resolution mesh models of individual sculptural panels at 0.5–2 mm point spacing, suitable for iconographic study and condition monitoring.

Common Mistakes

Conflating model types — producing a photogrammetric mesh when the client needs an HBIM, or producing an HBIM when the client needs a photorealistic render — is the most expensive mistake in heritage 3D modelling. Establish the deliverable format and its purpose before beginning any modelling work. A conservation architect requesting an HBIM for repair scheduling has no use for a beautifully rendered Blender model; a client requesting a visualisation for public display has no use for a Revit model they cannot open.

Failing to validate model accuracy against survey control data is a professional oversight. Every heritage 3D model should include a documented accuracy check: dimensions measured from the model compared against total station or tape measurements taken on site. A photogrammetric model that is beautiful but 5% too large is not a heritage documentation product — it is a 3D artwork. Report measured accuracy alongside the model as a standard deliverable.

Not documenting the model's survey date and condition state is a documentation failure. A 3D model of a heritage building is a record of how the building looked at a specific moment. Without a clearly documented survey date, photogrammetric models from 2026 and 2036 cannot be compared for change detection. Include survey date, model type, accuracy, software versions and coordinate reference system in metadata attached to every heritage model file.