looming op• Datenarten: signal → signal
• Aufruf: import fullseye as fs; fs.ledger.fly_tau_from_expansion(theta_signal, dt_s, shape='sphere') (die Implementierung direkt: import flyvision; flyvision.fly_tau_from_expansion(theta_signal, dt_s, shape='sphere'); aus dem Register: opsflyvision.get("fly_tau_from_expansion"))
> Für diesen Operator gibt es noch keine Übersetzung. Es folgt der Originaltext unverändert.
Time-to-contact from optical expansion — the tau margin.
From the subtended angle and its rate::
shape="disk": tau = sin(theta) / theta'
shape="sphere": tau = 2*tan(theta/2) / theta'
`theta_signal` is the full subtended angle over time, radians.
★ The two object models differ by 33% at a 60-degree subtense (`sin 60 = 0.866`
vs `2 tan 30 = 1.155`); using the wrong one silently mis-times a landing, so
the model is a required, named choice rather than a default guess.
theta_signal: the full subtended angle per sample, radians.
dt_s: the sample interval, seconds.
shape: `"disk" (a frontal circular disk) or "sphere"`.
Returns a 1-D float64 array of the time-to-contact per sample, seconds.
Non-expanding samples (`theta' <= 0) return NaN` — a documented
non-finite, because a contracting or static angle has no time-to-contact and
inventing one would be a plausible-wrong number. (The registry op
`tb_fly_tau_from_expansion` must return a finite signal, so it replaces those
NaN by the signal fallback without recording a fallback event —
`backend_safe.NONFINITE_BY_DESIGN`; call this function directly to keep the NaN.)
Ground truth: for the model's own object geometry (`theta = 2 asin(l/d)` for a
sphere, `theta = 2 atan(l/d) for a disk) approaching at speed |v|`, the
returned tau equals the true distance-over-speed `d/|v|` (pinned in the
tests).
Raises `ValueError`: a non-1-D / empty / too-short (< 3) / non-finite
*theta_signal*, a signal over :data:MAX_SIGNAL_POINTS, a non-positive *dt_s*,
and an unknown *shape*. A non-expanding angle is not an error — it is the
documented `NaN` return above.
• Katalog der Beispieldaten (Download-URLs / Lizenzen) — 2-D nutzt skimage.data (BSD/Public Domain) plus synthetische Bilder, 3-D nennt Download-URLs echter Datenquellen (Stanford, PDS, …).
• Herkunft und Literatur der Operatoren — die Quellen der Forschung/Verfahren, auf denen diese Operatorfamilie beruht.
• Der kanonische Algorithmus (Autor, Jahr) und seine Anwendungen stehen im Familienleitfaden oben.
• poc_fly_vision — py -3.11 examples/poc_fly_vision.py
signal als Eingabe)fly_emd_response · fly_lgmd_eta · fly_dsi
looming)*Provenance: flyvision.py — FLYVISION Operator-Registry. Diese Notiz wird von tools/opdocs.py md erzeugt (nicht von Hand bearbeiten).*
© 2026 Kazufumi Furuse — Fullseye operator documentation. Licensed under Apache-2.0.