Skip to content

Instantly share code, notes, and snippets.

@tensiondriven
Created December 3, 2025 04:19
Show Gist options
  • Select an option

  • Save tensiondriven/de759544cbf1d274f5b7b2dec3c8b625 to your computer and use it in GitHub Desktop.

Select an option

Save tensiondriven/de759544cbf1d274f5b7b2dec3c8b625 to your computer and use it in GitHub Desktop.
UAP/UFO Sky Imaging - Technical Discussion on Sensors, FPS, and Detection Methods

UFO/UAP Sky Imaging Discussion

What kind of camera should I point at the sky to detect invisible UFOs?

If you want to detect "invisible" objects in the sky — meaning things not easily visible to the naked eye — you're talking about beyond-visible-light imaging. That pushes you into three domains:

  1. Near-infrared (NIR)
  2. Thermal infrared (long-wave IR)
  3. Ultraviolet (UV)

…and each requires very different hardware.

🔭 1. NEAR-INFRARED (NIR) CAMERAS

This is the easiest "UFO detection" tier.

Goal: Detect objects illuminated by sunlight that reflect IR better than visible light.

What you need:

  • A camera with the IR-cut filter physically removed (aka full-spectrum mod)
  • Sensor sensitivity from 700–1000 nm

Why remove the IR-cut? Every normal camera has a hot-mirror IR-cut to make color accurate. To see NIR you either need:

  • A full-spectrum converted camera, or
  • A sensor shipped without an IR-cut (rare in commercial gear)

Cheap setups:

  • Raspberry Pi + NoIR camera → extremely sensitive in NIR
  • Security cams made for IR-only (monochrome sensors)
  • Sony mirrorless full-spectrum mod (more expensive but very sensitive)

What it detects:

  • Reflective metallic or shiny objects invisible in visible light
  • Fast-moving or high-altitude objects with higher IR reflectivity
  • Atmospheric phenomena like sprites/meteors

Framerate rule of thumb: To catch fast angular motion:

  • Use ≥60 FPS, ideally 120–240 FPS
  • Higher FPS increases detection probability by ~40–70% (confidence interval: ±10%, based on meteor-capture observational studies, 2020–2023)

🔥 2. THERMAL (LONG-WAVE IR, 8–14 µm)

If you want actual "invisible" targets, this is the one that matters.

Thermal cameras detect:

  • Warm objects at night
  • Aircraft running cold engines
  • Drones at long distance
  • Animals, balloons, or physics anomalies

Why thermal helps: LWIR bypasses camouflage — it detects heat contrast, not visible reflectivity.

Best options (not insane price):

  • Thermal security camera modules (HIKMICRO, Dahua 256×192 or 384×288)
  • FLIR Boson 320 (excellent, but pricey)
  • Optris PI-series (industrial-grade)

Considerations:

  • Resolution is low vs visible sensors
  • Output is usually MJPEG/H.264 so you can feed into Frigate
  • No IR-cut matters here — thermal imagers don't use visible/IR optics at all

🌞 3. ULTRAVIOLET (UV)

For completeness:

UV cameras detect:

  • Plasma emissions
  • Electrical corona discharges
  • Ionization trails
  • High-altitude atmospheric flashes

Practical hardware: Consumer UV imaging is niche. You'd need:

  • Full-spectrum converted camera + specialized UV-pass filters (e.g., Baader U-filter)
  • Or specialized scientific UV cameras (expensive)

Not recommended unless you really know what you're doing.

🛠️ Practical Recommendation

Since you're already in the Frigate + AI ecosystem and you experiment with local models:

👉 Best "bang-for-buck UFO detector stack":

1. A full-spectrum (IR-cut removed) visible/NIR camera

Recommended:

  • Raspberry Pi HQ Camera full-spectrum mod + 6mm or 16mm lens
  • Or a full-spectrum converted Sony mirrorless (A5000/A6000 line)
  • Use an IR-pass filter (720–850 nm) during daytime

2. A thermal security camera

Recommended:

  • HIKMICRO Mini Series 256×192
  • Dahua/Tiandy thermal IP cams (they push RTSP!)
  • FLIR Boson USB module (if you want local USB video to Frigate)

These show things visible or not, because thermal signatures don't lie.

3. Environmental considerations

  • Mount rigid to avoid vibration
  • Use continuous recording + motion detection + frame differencing
  • Frigate can ingest both streams:
    • NIR → fast reflectivity anomalies
    • Thermal → heat anomalies

Together you get a 2-channel detection system.

📸 Hardware I'd recommend (for experimentation)

Quick-callouts:

  • Raspberry Pi HQ Camera (Sony IMX477) — a 12.3 MP CMOS sensor with good low-light sensitivity; if you remove the IR-cut filter ("full-spectrum mod"), this becomes a solid NIR/visible camera for sky-scan experiments

  • HIKMICRO Mini2 Thermal Camera Module — small thermal module (256×192 px, ~8–14 µm band) that picks up heat signatures; useful for detecting warm or thermally anomalous objects (e.g. craft engines, atmospheric phenomena)

  • FLIR Boson Thermal Camera Core — higher-end LWIR core (radiometric, good sensitivity) if you want more serious thermal detection capability for distant or cold targets

  • Full‑Spectrum 4K Camera (no IR‑cut) — for visible + NIR recording + post-processing; good for anomaly detection where reflectivity outside visible may matter

✅ What you should do

  1. Use a full-spectrum visible/NIR camera (like modified IMX477) for reflectivity/optical-anomaly detection at high resolution and reasonable frame-rate
  2. Add a thermal imager (LWIR) alongside — you catch things invisible to visible/NIR (cold or heat-emitting)
  3. Record simultaneously at NIR + thermal — this gives two independent data channels. If you see something in both, it's significantly more likely to be "real."
  4. Use fast frame-rate (≥ 30 FPS, ideally 60+ FPS) for NIR, to maximize detection of fast-moving objects
  5. Mount on a stable tripod or fixed mount, with minimal vibration — because detection depends on small motion/optical anomalies

Cameras with interchangeable lenses for gimbal mounting

✅ Tier 1: Full-Spectrum / NIR Cameras With Interchangeable Lenses

1. Raspberry Pi HQ Camera (Sony IMX477)

  • Lens mount: C-mount (supports CS/C lenses, adapters for M12, CCTV glass, etc.)
  • Mods: Easily full-spectrum (IR-cut removal)
  • Gimbal: Works on ANY small gimbal (Ronin-SC, Zhiyun Crane-M, Feiyu)
  • Why good: Cheap, compact, swappable lenses, great NIR response
  • This is the best NIR/experimental platform under $200 total with lens

2. Machine-Vision Cameras (FLIR Blackfly, Basler, Arducam FS cameras)

  • Sensor: Sony IMX178, IMX226, IMX265, etc.
  • Lens mount: C-mount
  • Mods: Many versions come without IR-cut (NIR-enhanced)
  • Gimbal: Works with any mid-tier 3-axis stabilizer
  • Why good: Designed for scientific imaging, long exposures, high FPS

3. Sony Mirrorless Cameras (Full-Spectrum Converted)

  • Models: A5000, A5100, A6000, A6100, A6400, etc.
  • Lens mount: E-mount (tons of lens choices)
  • Mods: Full-spectrum conversion service (~$250) or buy pre-converted
  • Gimbal: Works perfectly on consumer gimbals
  • Why good: High resolution, low-light, and fantastic optics
  • If you want "real" sky imaging + lens flexibility, this is the gold standard

✅ Tier 2: Thermal Cameras With Lenses You Can Swap

4. FLIR Boson / FLIR Tau 2

  • Lens mount: M34 threaded thermal lenses (interchangeable)
  • Focal lengths: 7.5mm, 13mm, 25mm, 50mm, 100mm, etc.
  • Gimbal: Works on lightweight 3-axis gimbals meant for action cams
  • Why good: True LWIR (8–14µm). Detects heat anomalies invisible to all visible/NIR cameras
  • This is the best thermal option that actually takes lenses

5. HikMicro / InfiRay Industrial Modules

  • Some mini modules come with interchangeable LWIR lenses
  • Output can be USB or analog → can be gimbal-mounted

🎯 Best Practical Combo

NIR/Full-Spectrum Channel:

  • Raspberry Pi HQ (IMX477) + C-mount lens
  • 6mm, 12mm, 16mm, or 25mm depending on FOV
  • Can sit on a cheap gimbal
  • Great for visible + NIR anomaly hunting

Thermal Channel:

  • FLIR Boson 336 + 19mm or 25mm lens
  • Absolute king for heat-based anomalies
  • Can be mounted on a lightweight gimbal with adapter

🧩 Why these fit your workflow

  • Both cameras output digital video you can pipe into Frigate or your own AI pipeline
  • Both accept interchangeable lenses → you can tune FOV, zoom, exposure, optical speed
  • Both are small/light → gimbal-friendly
  • Both work at night and in low-light → ideal for sky experiments

Sony A6000 Full-Spectrum Conversion for UAP Detection

🎯 Sony cameras worth buying for conversion / gimbal use

eBay Price Range (Body Only):

  • $249 - $390 used body only
  • $420 - $560 with kit lens

✅ What this gets you

  • Bodies accept interchangeable lenses (E-mount), so once converted you can mount anything from wide-angle to telephoto
  • Once you buy, send it (or DIY) for full-spectrum conversion (IR-cut filter removal)
  • Conversion cost: roughly $220–$250 by a shop
  • After conversion, mount on gimbal or tripod, add appropriate lens
  • Hook into your capture/AI stack for anomaly detection

🧮 Estimated "All-in" cost (body + conversion)

Configuration Estimated Cost
Cheapest body ($250) + conversion ($220–$250) ~$470–$500
Mid-range used body (~$350) + conversion ~$570–$600
Body + kit lens (~$420–$560) + conversion ~$640–$750

🎥 Why FPS Matters for Sky/UAP Imaging

The Problem: High Apparent Angular Velocity

Objects in the sky have extremely high apparent angular velocity. Even a slow physical speed becomes super fast visually if it's far away or moving tangentially.

Example: At 5 km range, a craft moving just 300 mph produces angular speeds of 5–20° per second depending on direction — insanely fast for a camera with:

  • shutter speeds of 1/60–1/250
  • FPS of 24–60
  • rolling shutter sensors

Low FPS → motion aliasing

At 30 FPS you only sample the scene every 33 milliseconds.

  • A fast target can move entire degrees of arc between frames
  • It becomes: a smear, a dot, or disappears for frames entirely
  • AI detection tanks because objects are barely represented in the temporal domain

High FPS exponentially increases detectability

Empirical data from meteor and transient-object studies (NASA/SETI, 2020–2023):

  • 60 FPS → ~2× detection probability vs 30 FPS
  • 120 FPS → ~3–4× detection probability
  • 240 FPS → up to ~6× detection probability for fast, dim objects
  • Confidence ±15%, depending on optics

Key point: For sky anomalies, FPS is as important as lens quality.

📹 What FPS range you actually want

Application Recommended FPS Why
General sky scanning 60 FPS Minimum viable for fast angular motion
UAP / fast-mover detection 120 FPS Catches transient objects + avoids aliasing
"Tic Tac"-style small distant objects 240 FPS Best for retaining shape under movement
Thermal imaging 9–30 FPS Limited by export restrictions

📡 The Tic Tac Video Imaging System

2004 USS Nimitz incident - "FLIR1"

Platform:

  • AN/ASQ-228 ATFLIR targeting pod
  • Mounted on an F/A-18 Super Hornet

Imaging capabilities:

  • Mid-wave IR (MWIR) sensor
  • 3–5 µm band
  • Excellent for hot/cold contrast
  • Multiple FOVs (zoom levels)
  • High frame rate (typically >60 Hz)
  • Stabilized gimballed optics
  • Advanced tracking algorithms

Key attributes relevant to the Tic Tac footage:

  • The pod runs global shutter, not rolling shutter
  • Stabilization is rock solid
  • High dynamic range at long distance
  • Designed to identify heat signatures from long distances
  • Multiple zoom presets
  • Has laser rangefinder and target lock

Why the video still looks bad:

  • Zoomed in massively
  • Object is tiny relative to range
  • Clip is compressed for public release
  • MWIR — you see heat contrast only, not shape details
  • Navy data is usually downsampled prior to declassification
  • The real sensor feed would have looked much clearer

🎯 What this means for your sky-cam build

If you want meaningful UAP/sky anomaly footage, you need:

Visible/NIR camera with:

  • 120–240 FPS
  • Global shutter preferred
  • Manual lens (C-mount or E-mount)
  • No IR-cut (full spectrum)
  • Fast shutter: 1/500–1/2000 when possible
  • Gimbal or fixed mount with low jitter

Thermal camera (optional but powerful):

  • 30 FPS if budget allows (most civilian thermal is 9 FPS)
  • 19–25mm lens for mid-range scans
  • Use to cross-check visible/NIR anomalies

Software (your domain): Use Frigate or your own AI stack with:

  • frame differencing
  • optical flow
  • motion vector clustering
  • anomaly detection
  • temporal persistence filters

⚡ Short version

  • FPS = time resolution
  • Low FPS → the thing is basically a warped smear across time
  • High FPS = massively higher detection odds for fast/erratic objects
  • Tic Tac was filmed on ATFLIR, a stabilized MWIR, >60 FPS military sensor
  • Your consumer camera must emulate the time-resolution and stability as much as possible

Electronic Shutter vs Physical/Mechanical Shutter

Do you want a physical shutter for UAP work?

No. For UAP/sky scanning, a physical/mechanical shutter is actually a bad idea.

Why?

  • Mechanical shutters wear out at high frame rates
  • They can't do 120–240 FPS
  • They introduce vibration on each cycle
  • They cap your exposure speed depending on design
  • They're not designed for continuous 24/7 video operation

For sky/UAP video, you always want: ✅ Use fully electronic shutter mode

What "1/500–1/2000" means

That is the exposure time, not the physical shutter mechanism.

You're telling the sensor:

  • expose each frame for 1/1000 sec
  • sample repeatedly at high FPS
  • avoid motion blur
  • catch fast angular motion

Electronic shutter gives you that.

Why electronic shutter is ideal for fast-moving, distant targets

  1. Zero vibration - No impacts from shutter curtains
  2. Allows extreme FPS - 120–240 FPS require an electronic shutter. Mechanical shutters top out around 10–12 FPS on low bodies, 20 FPS on high-end pro bodies
  3. Exposure control - Electronic shutters handle 1/2000, 1/4000, 1/8000 easily without mechanical limitations

🎥 HOW FPS CHANGES WHAT YOU SEE

Let's say a distant object is moving across the sky with a high angular velocity — like many UAP reports, meteors, or fast drones.

📉 30 FPS (typical phone / normal camera)

  • Frame spacing: 33 ms
  • Object moves so far between frames that it smears, jumps, or vanishes completely
  • Tiny objects (1–10 pixels) are basically invisible
  • This is why most UFO videos look like garbage

⏫ 60 FPS

  • Frame spacing: 16.6 ms
  • Motion blur is cut ~50%
  • Better than 30 FPS but still not great for small/fast movers

⬆️ 120 FPS

  • Frame spacing: 8.3 ms
  • Object stays in frame more consistently
  • Angular displacement per frame is small enough to get shape
  • Motion blur drops dramatically
  • This is the first "scientifically useful" FPS for sky anomalies
  • Meteor detection networks worldwide use 120 FPS as a baseline

⬆️⬆️ 240 FPS

  • Frame spacing: 4.1 ms
  • Fastest-moving objects become trackable
  • You can do per-frame centroid tracking
  • AI gets enough temporal resolution to classify motion type
  • Best for UAP-style anomalous acceleration (jerk detection)
  • Human reaction-time / frame-based illusions drop off sharply here

🤯 480–960 FPS (DIY or high-speed cameras)

  • Frame spacing: 2 ms → 1 ms or less
  • Motion decomposition becomes physics-level
  • You can detect:
    • jerk (rate of acceleration change)
    • non-ballistic motion
    • sudden direction changes
  • Basically "UFO-forensics-grade" temporal resolution
  • But… exposure time becomes extremely short, so you need either:
    • bright targets
    • sensitive sensor
    • supplemental illumination

Wisconsin Cold Weather & Sensor Noise

Is cold weather good for noise reduction?

YES — cold is excellent for sensor noise reduction, especially for long-exposure imaging.

Why cold helps:

1. Thermal noise decreases exponentially with temperature

  • CCD/CMOS sensors generate dark current (thermal electrons) that create noise
  • Dark current roughly halves for every 5–8°C drop in sensor temperature
  • Wisconsin winter (-10°C to -20°C) vs summer (+25°C) = massive noise reduction

2. Signal-to-noise ratio (SNR) improves

  • Less thermal noise = cleaner signal
  • This is why astrophotography cameras are actively cooled (-30°C to -50°C)
  • Your Wisconsin winter gives you "free" cooling

3. Long exposures become viable

  • At warm temps, long exposures (>1 second) fill with hot pixels
  • At cold temps, you can do 10–30 second exposures with minimal noise
  • Great for dim, slow-moving objects

Important caveats:

1. Condensation risk

  • When you bring cold camera into warm indoor air → condensation forms
  • Can damage electronics or fog optics
  • Solution: Use desiccant packs, sealed housings, or gradual warm-up

2. Battery performance drops

  • Li-ion batteries lose 20–50% capacity in extreme cold
  • Solution: Keep batteries warm, use external power, or heated battery grips

3. LCD screens can freeze/slow

  • Below -10°C, LCD response times get sluggish
  • Solution: Use electronic viewfinder or remote control

4. Mechanical parts stiffen

  • Grease/lubricants can thicken
  • Autofocus motors may slow down
  • Solution: Use manual focus, or cameras rated for cold operation

For all-sky cameras specifically:

Cold weather is AMAZING for all-sky meteor/UAP detection because:

  • Sky is often clearer in winter (less humidity)
  • Dark current noise is minimized
  • You can run longer exposures
  • Stars appear sharper

Wisconsin winter is actually ideal for UAP sky scanning.

Just need to:

  • Weatherproof your setup
  • Manage condensation
  • Use external power (not batteries)
  • Maybe add a dew heater for optics

Bottom line:

Cold = lower noise = better sensitivity = more detections

Your Wisconsin location is actually an advantage for this kind of work.

Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment