Skip to content

Confirm or withhold the focal-120 fallback row (11,984 frames on an unverified camera identity) #40

Description

@NewGraphEnvironment

Problem

inst/extdata/camera_formats.csv ships one row whose warrant is an identity argument
rather than a measurement, and it covers 11,984 frames — 5.4% of the digital
catalogue.

The "120" fallback row (key_type = focal_length) gives catalogue-focal-120 frames the
Intergraph DMC's 165.888 x 92.160 mm format. No calibrated camera in the record has focal
120: the value is reached by extrapolating from focal 127, and the argument for it is that
the only large-format 120 mm body in the provincial record is the Z/I DMC, whose own
report gives a virtual focal of exactly 120 mm — while the catalogue records 127 for those
same frames.

So the row is probably right, and it is right for a reason the generator cannot check.

Why this is worth a decision rather than leaving as-is

#32 refused the focal-83 fallback on exactly this kind of reasoning being wrong: the
nearest calibrated focal was 3.6% away and the true camera was a medium-format body
93% narrower. Extrapolation distance does not predict the error — focal 120 reaches
further (5.5%) and is fine. Only camera identity separates them, and that is precisely
what is unverified here.

The row is marked extrapolated = TRUE with width_spread_pct = NA, so it is
filterable and does not claim confidence it lacks. That is why it shipped rather than
blocking #32.

Options

  1. Leave it. 11,984 frames get a footprint that is probably correct, flagged as
    inferred. Consistent with "rather a poor outcome than nothing".
  2. Confirm it. The focal-120 frames are all 2011. OPERATION_TAG / FILM_ROLL or a
    flight log via fly_fetch(type = "flight_log") may name the operator or camera; a
    CAMERA_CALIBRATION_URL may exist on a neighbouring roll from the same operation.
    If confirmed, the row stops being an extrapolation.
  3. Withhold it. Those frames return to unknown_format. Safest, and costs the
    largest single block of inferable frames in the table.

A cheap partial check exists and needs no fieldwork: the implied-ground-elevation test
(check F in data-raw/make_camera_formats.R) applies to any frame with a populated GSD
and FLYING_HEIGHT. If the focal-120 frames have those, a 165.888 mm format either
implies a plausible BC elevation or it does not.

Value

If we do it: either 11,984 frames stop resting on an unverified identity, or they are
withheld deliberately rather than by default. Both beat the current state, which is a
judgement call nobody has revisited.

If we never do: the row stays flagged and filterable, and a user who cares can drop
it with footprint_basis != "inferred_format". The cost is that the one number in the
table with no measurement behind it looks like the other 18.

Follows #32

Metadata

Metadata

Assignees

No one assigned

    Labels

    No labels
    No labels

    Projects

    No projects

      Milestone

      No milestone

      Relationships

      None yet

      Development

      No branches or pull requests

      Issue actions