Why it matters
Without this, misreads of pigment stability, hidden damage remains, condition maps stray from actual states, and repairs misalign with original manufacture, inviting later deterioration, project overruns, and compliance gaps.
From design to safety
Design
AI guides inspection protocol design
The system reviews data schemas, imaging modalities, sampling plans, and external cataloging standards to establish a standardized workflow for multispectral scans and subsequent condition mapping. It prescribes sensor calibration intervals, lighting angles, colour targets, and calibration artifacts to ensure reproducible data across sessions and items, including documented tolerances for colour accuracy and spectral alignment.
- UV fluorescence
- IR imaging
- image sets
Build
Acquire and collate spectral data
It reads calibration files, spectral bands, panel references, and item metadata to validate scan quality, ensures cross-spectral alignment, and assembles a cohesive condition map per item with per-page granularity and bound leaves linkage, incorporating ESTC identifiers, provenance notes, and binding history in the index to support future re-evaluations.
- raking light
- transmitted light
- condition maps
- image sets
Operation
Run scans and generate condition outputs
The AI processes live scans, aligns bands, detects damage patterns, and updates item records with defect localization and severity, while creating per-page overlays and exportable XML or JSON condition bundles for lenders and insurers, plus a change-log referencing conservation actions and recommended treatment priorities.
- raking light
- UV fluorescence
- transmitted light
- infrared
Maintenance
Review drift and revalidate maps
Periodic re-imaging checks for sensor drift, recalibration needs, and re-stratification of condition overlays against baseline maps, with diffs saved to a versioned archive and alerting for unresolved conflicts, including quarterly audits of pigment stability indicators and paper degradation indices.
- image sets
- condition maps
Safety
Ensure provenance and compliance integrity
The system logs imaging provenance, access controls, and sign-offs by conservators to support audits and legal compliance, storing tamper-evident records and chain-of-custody notes tied to each item, with secure timestamping and export controls for sensitive holdings.
- UV fluorescence
- transmitted light
- infrared
Part by part
- 1
raking light
- Watch for
- Spaces of abrasions or pigment loss appearing only under oblique illumination.
- AI action
- Detect depth of relief changes and flag high-risk regions for conservator review.
- 2
UV fluorescence
- Watch for
- Fluorescent varnish or overpaint masking underlying paper damage or foxing patterns.
- AI action
- Highlight areas with anomalous fluorescence for targeted testing.
- 3
transmitted light
- Watch for
- Brittle page sections or differential staining revealed by backlighting.
- AI action
- Annotate layer boundaries and guide micro-sampling decisions.
- 4
infrared
- Watch for
- Late-stage binder changes or text obscured by varnish; detect ink fading.
- AI action
- Flag contrast shifts and generate monochrome overlays for conservation planning.
- 5
image sets
- Watch for
- Misregistration or uneven exposure across spectral bands.
- AI action
- Compute coregistration confidence and alert for re-scans.
- 6
condition maps
- Watch for
- Outdated or incomplete condition overlays not aligned with current stock.
- AI action
- Publish updated condition maps with per-item change logs.
Worked example
Catching a faint ink bleed before a binding repair
Illustrative scenario with sample data.
A mid-18th-century binding requires reattachment while preserving ink fidelity; multispectral imaging assesses pigment stability and bleed risk.
Signal in
stockID: 482-GL, itemType: binding, scanDate: 2026-03-12, channels: [raking, uv, transmitted, infrared], baselineMap: CM-482-03
System out
alert: bleed-prone area detected at folio 42r; recommended restraint method and re-scan plan; condition map updated to include new defect regions (BS-042-bleed-ML).
Human sign-off: Conservator reviews the alert, approves the restrained repair plan, and authorizes micro-sampling if needed.
How to roll it out
Phase 1
Weeks 1 to 4
Baseline imaging workflow defined, calibration protocols established, and initial item cohort prepared.
Phase 2
Weeks 5 to 12
Pilot scans expanded, condition maps integrated with stock records, and staff training completed.
Phase 3
Weeks 13 to 20
Full deployment with ongoing monitoring, maintenance checks, and archival of imaging provenance.
Where people stay in charge
- Only qualified conservators or conservators-in-charge may approve sampling and repair recommendations based on AI outputs.
- All imaging data and condition maps must be signed off by the lead conservator and logged for provenance audits.
- Access to raw spectral data and edited maps is restricted to authorized personnel only.
Questions owners and crews ask
What is multispectral imaging in conservation and how does AI help?
Multispectral imaging captures data beyond visible light; AI analyzes spectra, flags anomalies, and translates data into actionable conservation decisions such as treatment planning and documentation.
Which items should get AI multispectral scans first in a conservation program?
Prioritize items with fragile pigments, unknown bindings, or evident in-painting; start with a small cohort to validate protocols before scaling.
How does imaging protect provenance and archival records?
Imaging logs provide traceable evidence of condition and modifications over time, supporting provenance checks and legal compliance through immutable records.
What training is needed for conservators to use these tools?
Conservators require protocols for calibration, data interpretation, ethical handling of sensitive images, and sign-off procedures to ensure correct application.