Sensofusion Drone Surveillance

Sensofusion Drone Surveillance turns your phone into a portable drone-spotting
sensor. It runs neural-network detection entirely on the device, plots
sightings on a map, and — when multiple phones are connected to the same
Sensofusion network — triangulates the actual position of the drone in 3D from
the bearings each device contributes.

WHAT IT DOES

• Live drone detection on the camera feed using an on-device ML model
(drone_detector_v3) running through ONNX Runtime, accelerated by NNAPI on
supported hardware.
• Three selectable detection backends: a fast heuristic detector for low-end
devices, the neural ONNX model for accurate tracking, and a TensorFlow Lite
path for comparison and benchmarking.
• Real-time bounding boxes that follow the drone across the frame, with a
motion tracker that smooths jitter and bridges short occlusions.
• Compass and motion-sensor integration to record the precise bearing of every
sighting, and GPS-tagged sighting reports so the bearings can be combined
into a global position fix.
• Audio analysis that listens for the propeller signature of small multirotor
drones, widening the detection cone beyond the camera’s field of view.
• Simulation mode that renders a synthetic sky natively and runs the real
detectors against it — useful for testing your settings, demoing the app,
and verifying the model still fires after a software update without needing
a drone in the air.
• Cross-device fusion: when several phones running the app are signed in to
the same Sensofusion deployment, the backend combines their sightings into
triangulated 3D tracks and overlays nearby ADS-B aircraft for context.

WHO IT IS FOR

Sensofusion Oy builds counter-UAS and drone-awareness systems for critical
infrastructure operators, event security teams, and law-enforcement units.
This app is the field-portable extension of the Sensofusion ground-based
sensor network: a way for people on site to add visual confirmation and
extra bearings without carrying a dedicated sensor.

PERMISSIONS

• Camera — required to run the visual drone detector on the live feed.
• Microphone — optional, used by the audio detector to listen for drone
propeller signatures.
• Location (fine + coarse) — used to geo-tag sightings so multiple phones
can be combined into a triangulated track.
• Network — to talk to your Sensofusion backend; nothing is sent off-device
unless you sign in to a deployment.

PRIVACY

All ML inference happens on the device. Camera frames are not uploaded.
Sighting reports (timestamp + GPS + bearing + a small thumbnail) are only
sent to the Sensofusion backend you configure; there is no third-party
analytics SDK in the build.