Collect
Registry
Every source documented: type, location, access date, specs, and what you extracted from it.
Provenance
Who provided it, which repository, original URL — the chain of custody begins here.
Characterize
Assess optics, artifacts, and metadata to decide whether and how each source can be used.
This is Stage 1 of the Research Gate. Collection is where you acquire sources and document them — it is read-only characterization, not transformation. The dividing rule: if you are recording or deciding about a source, it belongs here; if you are modifying it, that belongs to process.
The Source Registry
Create a record of every piece of evidence you use. This establishes what the Berkeley Protocol calls "chain of custody": the chronological documentation showing where evidence came from, how it was obtained, and that it hasn't been altered.
A simple spreadsheet or document listing all sources — with columns for: source name, type, location, date accessed, what information was extracted, reliability assessment.
For each source, document:
- What it is: File name, document title, archive location
- Where it came from: Who provided it, which repository, URL with access date
- Technical specifications: Resolution, format, equipment used to capture it, date created
- Verification: How you confirmed it's authentic and unmanipulated (for key evidence, include file hashes like SHA-256)
- What you used it for: Which measurements, which references, which parts of your model
Source Documentation Template
SOURCE: [Name/title]
TYPE: [Photograph/Drawing/Measurement/Document/Video]
LOCATION: [Archive/URL/Physical location]
ACCESS DATE: [When you obtained it]
FILE HASH: [SHA-256 if applicable for digital evidence]
SPECIFICATIONS: [Resolution/dimensions/format/etc.]
USED FOR: [What part of model/what information]
RELIABILITY: [High/Medium/Low - why?]
NOTES: [Limitations, concerns, context]
Scale this to your project:
Simple historical reconstruction: "Floor plan dimensions derived from Smith, Architecture of Colonial Boston (1982), Figure 12, p. 47, cross-referenced with photograph dated 1898 from Boston Public Library Digital Collection (image BPL_1898_047, accessed May 2024)."
Forensic reconstruction: Include file hashes, metadata analysis, custodial information, equipment calibration dates, measurement protocols.
Provenance
Provenance is the documented history of where each source came from. Record:
- Who provided it: the individual, organization, or repository.
- How it was obtained: direct download, forwarded file, official request, scraping.
- Original location: the canonical URL or archive path (not just your local copy).
- Date obtained: when you acquired it.
This is the foundation of the chain of custody that continues in preserve-store (hashing, working copies) and is reported in validate-report.
Authenticity Gate
For key evidence, you need to demonstrate how you confirmed the source is what it claims to be and has not been manipulated before you received it.
Ask, and document the answer:
- Is this the canonical record? Did it come from an official source, or a relaying third party?
- Are there signs of fabrication? Metadata mismatches, compression anomalies, editorial artifacts, logical impossibilities.
- Does it corroborate? Does it agree with other independent sources about the same facts?
- Can it be authenticated? For critical evidence you should be able to answer how — file hash from the source, signed metadata, official registry, direct physical custody.
If you cannot authenticate a source, flag it. Flagging is not failure — it is honest methodology. An unauthenticated source can still be used as supporting context, but it must be labelled as such and weighted accordingly.
Characterize Sources (Technical Assessment)
Before deciding how to use a source, characterize its technical properties. This informs processing decisions (in process) but is itself read-only analysis.
Optics: Analyze footage for lens distortion types (Barrel, Pincushion, Mustache) and vignetting per main source file.
Artifacts: Identify motion artifacts (Motion Blur, Rolling Shutter, Compression blocks).
Geometry: Identify sources with optimal perspective lines (One-point, Two-point, or Three-point) to determine the best solving method for each specific angle.
Relevance confirmation: Each source must demonstrably support the research question. Flag anything collected "just in case" — either give it a stated purpose or note it as contextual.
Metadata Extraction
Action: Analyze file headers (EXIF/XMP) for each main source file (look at the main video, not videos derived from it).
Key data: Look for camera make/model, focal length, sensor size, date/time, and GPS coordinates.
Note: If the source is from social media, metadata is likely stripped; rely on visual "Technical Assessment" instead.
If no metadata: Estimate whether each specific camera uses a Prime Lens (Fixed) or Zoom Lens (Variable focal length). Estimate the focal length for fixed lenses (wide-angle shots, zoomed shots) and possible sensor width (mm) for each profile independently. Document these estimates — they become inputs to your camera-parameter assumptions, and any grid of assumptions must be visible.
What Goes Where (guard rails)
| Concern | Stage |
|---|---|
| Recording what a source is, where it came from, its specs | collect (this page) |
| Characterizing suitability (optics, artifacts, perspective, metadata) | collect (this page) |
| Protecting originals, hashing, working copies, folder structure | preserve-store |
| Splitting clips, extracting frames, enhancement, temporal alignment | process |
| Ground truth, key objects, uncertainty, decision log | analyze |
Next stage
Sources collected and documented → Preserve & Store: protect originals, hash everything, and separate working copies.