The birdhouse project is based around a Nistkasten from Vogelwarte Sempbach and upgraded to capture live video and sensor data using a Raspberry Pi. It publishes the stream via mediaMTX, and serves telemetry through this website. Extended live diagnostics are available for nerds.
The first iteration was an all-in-one Django app running directly on the Raspberry Pi. The second iteration moved web delivery to a dedicated server for better reliability and scale. The original project is archived at github.com/hacknus/birdhouse-monitor.
A Raspberry Pi 3B handles camera capture, GPIO sensors, and control commands. The Raspberry Pi is connected to the web-server via a TCP connection.
The following components are mounted in the birdhouse:
We included a microphone to analyse the sounds of different birds. A spectrogram is shown on the website, where y-axis corresponds to the frequency (logarithmic) and the x-axis to time (full window is about 20s).
If a plane flies overhead, the noise floor increases, sometimes the Doppler effect can be visible. Emergency vehicles and rain drops can also be detected.
The camera does not have an infrared filter and is thus able to take images during the night when the birdhouse is illuminated with IR LEDs. Birds are not disturbed by infrared light, since they can see better in the UV/blue part of the spectrum than in the red/IR [1, 2].
To detect motion in the birdhouse, a 60 GHz radar has been mounted on the ceiling. It is not only able to detect motion but also provides an activity score and breathing rate (experimental). 60 GHz radiation is not dangerous to birds as it does not penetrate more than micrometers into the tissue [3].
The camera stream from the Raspberry Pi is distributed through mediaMTX which exposes HLS/WebRTC stream paths.
Live metrics and spectrogram data are rendered in parallel through Grafana and WebSocket-driven visualizations.
The webserver and Raspberry Pi are connected through an encrypted Netbird VPN.
The main software for the raspberry pi is written in Python and can be found on https://github.com/hacknus/birdhouse-monitor. The spectrogram streaming is written in Rust and can be found on https://github.com/hacknus/spectrogram-ws-rs. The software for the webserver is written in Rust with Dioxus and can be found on https://github.com/hacknus/birdhouse-rs.
