The Project How it works The Station What's next
The project

A personal
experiment

Lisbon Finals started as a simple question: could a Raspberry Pi on a terrace automatically photograph every plane that flies overhead — and post it to Instagram without any human input?

The answer, after months of iteration, is yes. What began as a weekend project became a full autonomous monitoring station — tracking noise levels, flight routes, airline patterns and atmospheric data, 24 hours a day, from a Lisbon terrace facing the LPPT approach path.

3 May 2026
Recording since
Flights tracked
Photos published
24/7
Autonomous operation
0
Human triggers per post

How it works

Three layers,
zero human triggers

01 📡
Detect & Measure

A small radio receiver picks up the signals every commercial aircraft is required to broadcast — giving us the flight number, altitude, speed and exact position in real time. A microphone simultaneously measures the noise level on the terrace. The system only triggers when both conditions are true: a plane is close enough, and loud enough.

RTL-SDR ADS-B 1090 MHz readsb dB SPL 5 km · 5,000 ft
02 👁
Capture & Score

A 4K camera continuously streams footage of the approach path. Every frame is instantly rated by an intelligent scoring system — only shots where a plane is clearly and sharply visible make the cut. The best frame is then paired with live flight data and sent to an AI to write the caption.

Reolink 4K RTSP YOLO v8 Claude Haiku 0.30 threshold
03
Enhance & Publish

The chosen photo is processed by an AI image enhancer that reconstructs fine detail at 4× the original resolution — about 15 minutes on the station's small onboard computer. It is then automatically scheduled, avoiding quiet hours at night (11pm–7am). Everything goes live on @lisbon.finals with the AI-written caption and flight data — no human involved.

Real-ESRGAN ×4 ONNX Runtime Cloudinary Meta Graph API Raspberry Pi 4
Full automation pipeline · end to end · zero human trigger
RTL-SDR
1090 MHz
ADS-B
decode
Acoustic
trigger
RTSP
capture
YOLO v8
scoring
Telegram
validation
Real-ESRGAN
4K upscale
Haiku
caption
Instagram
Graph API

The station

Hardware
in service

🖥
Raspberry Pi 4 — 2GB
Central brain — runs all services 24/7: ADS-B decode, capture trigger, YOLO scoring, scheduler, Flask admin server.
● EN SERVICE
📡
Nooelec NESDR Smart v5
RTL-SDR dongle receiving ADS-B signals at 1090 MHz. Replaced June 2026 after thermal failure — now fitted with heatsinks.
● EN SERVICE
📷
Reolink 4K PTZ TrackMix
4K camera streaming over RTSP. Aimed at the LPPT approach path — every frame analyzed by YOLO v8 in real time.
● EN SERVICE
🔊
GAIN EXPRESS SPL Meter + ADS1115
Professional sound level meter (30–130 dB). DC output wired to ADS1115 ADC → I²C → Pi. Tested on bench — outdoor deployment pending enclosure.
◎ TESTED · SOON IN SERVICE
🌬
GY-SGP30 Air Quality Sensor
eCO₂ and TVOC monitoring. I²C → Pi. Tested and calibrated — will go live with the outdoor enclosure.
◎ TESTED · SOON IN SERVICE
🌀
MPU-6050 Dual IMU — Vibration
Two accelerometers (AD0→GND + AD0→VCC, addresses 0x68/0x69) on the same I²C bus. Detects ground vibration from passing aircraft — a dimension beyond sound.
◎ TESTED · SOON IN SERVICE
🏭
PMS5003 — Fine Particle Sensor
Laser particle counter measuring PM1.0, PM2.5 and PM10 in real time. UART → Pi. Correlates each aircraft pass with a concrete air quality reading.
◎ TESTED · SOON IN SERVICE
📦
Outdoor Plexiglass Enclosure
Custom-cut 4mm PMMA box (20×40×30cm) with angled lid, ventilation holes, cable gland and internal shelf. Laser-cut at Fab Lab Lisbon.
◐ IN PROGRESS — June 2026

Upgrades
in progress

The station is continuously evolving. New hardware and sensors are being integrated to capture even more — from vibration analysis and industrial-grade acoustics to live ATC communications and air quality monitoring.

Support the project
All hardware is self-funded. If you find this project useful or just enjoy watching planes, a coffee goes a long way toward the next sensor.
Buy Me a Coffee at ko-fi.com
🔊
130 dB Sound Sensor

An industrial-grade sound meter will replace the current microphone. The new sensor handles up to 130 dB without maxing out — capturing the full impact of a low-flying aircraft as real numbers, not just a ceiling.

Hardware · Ordered
🌀
Dual IMU Vibration

Two motion sensors mounted on the terrace will detect the physical vibration caused by passing aircraft — the kind you feel in your chest but a microphone can't fully capture. A new dimension of impact, beyond sound.

Hardware · I²C
🌬
Air Quality Station

Two environmental sensors will measure fine particles and air pollutants over the terrace in real time — correlating each aircraft passage with a concrete air quality reading. A first for a private monitoring station in Lisbon.

Environmental · SGP30 · PMS5003
📡
Live ATC Listening

A second radio receiver tuned to Lisbon tower's frequency will capture live communications between pilots and air traffic control — linking each aircraft to the exact instructions they received in real time.

RTL-SDR · 118.1 MHz