Every band.
One picture.
Fused at the edge.
RF direction-finding, acoustic and seismic arrays, EO/IR, radar and covert emplaced nodes — every band classifies its own signal at the sensor and lands on one shared picture. A gunshot becomes a located impulse, a plate becomes a read, a footfall becomes a classed contact: ground truth for the areas you hold, and new hardware is an onboarding step, not a rebuild.
DETECTION BAND
DETECTION LEDGER · LIVE
Every field sensor on one timeline
The covert field runs many sensing modalities at once — EO/IR and thermal cameras, number-plate readers, acoustic gunshot arrays, buried seismic geophones, RF direction-finders and ground-surveillance radar — and they converge on one shared picture. Each modality keeps strict class ownership: the plate reader owns vehicles, the acoustic array owns impulsive events, the radar owns moving tracks — so one sensor never overwrites another’s call. A detection from any band is one observation among many, placed, timed and trusted on the same footing.
A gunshot heard on the acoustic array and a bearing taken by RF direction-finding cross to a shooter location in the same window — multi-modal, not single-sensor.
A number-plate read and a radar micro-Doppler track agree on the same vehicle, so identity and motion arrive together rather than as two unlinked feeds.
Object classes form an exhaustive, typed set — adding a class is a deliberate change, not silent feed drift.
Conditions are carried with every detection — a thermal camera at night, a geophone under canopy — so confidence reflects what each sensor could actually sense.
Ground truth from emplaced nodes
Covert emplaced nodes — cameras, seismic geophones, perimeter trip lines and low-power trackers — register as first-class field sensors with their own trusted identity. This modality detects its own subset of the scene: man-down, panic, tamper, jamming, low-battery, ignition, perimeter breach and ANPR plate reads, each resolved at the node before it leaves the field. The emplaced layer is the ground-truth complement to the overhead and RF bands — physical presence where imagery and tracks cannot see.
Each unit shows one clear status that always surfaces the most urgent condition first — life-threatening events take priority over jamming, then connectivity, then battery.
ANPR cameras emit plate reads, perimeter sensors emit breach events, seismic geophones emit footfall and vehicle signatures — each node stays inside its own detection class.
Position reports can carry battery, link, ignition and jamming status, so even basic legacy trackers integrate without new hardware.
A jamming detection is itself awareness: a node going dark under RF denial is a positive signal, not a gap.
Normalise once, at the sensor
At the sensor edge, every raw signal is reduced to one clear, typed detection before anything leaves the collection layer: a camera frame becomes a classified object, an acoustic impulse becomes a located gunshot, a geophone waveform becomes a classed footfall. This is where the physics ends and the shared picture begins — the sensor decides what it saw and stamps it with time, position, class and confidence, emitting a single fact per observation. Running normalisation at the edge means the picture ships already typed, not raw — bandwidth carries decisions, not pixels.
raw signal typed at the sensor — one clear fact per observation
raw signal becomes a classed fact
time · position · class · confidence
every event counted once downstream
BLUEPRINT · EDGE NODE
EMPLACED NODE · EXPLODED VIEW
DWG 03-C · NODE PACK
Normalisation runs as close to the sensor as the hardware allows: on-node for emplaced field sensors, in the camera for EO/IR, at the receiver for RF and radar.
The datastream pipeline keys on those stamped attributes downstream — nothing re-derives them.
Streaming sensors and polled sensors reduce through the same edge normaliser regardless of source format.
Spatial density is summarised at the sensor, so the picture arrives map-ready instead of as raw points.
SEE ALSO
Past the edge, every detection is counted exactly once — no duplicates — see OSINT Datastreams.
New hardware, one shared picture
Bringing a sensor online is a hardware-integration problem, not a software rebuild. A custom or integrated piece of hardware — a new covert node, an RF receiver, an acoustic array — is integrated once and then behaves like every other sensor. The registration declares which detection classes the sensor owns; from that point its detections flow on the same path as every other sensor. Integrating hardware is the only sensor-specific work — the picture it feeds is shared, not rebuilt per device.
Integration is a small, well-defined hardware step; everything downstream stays unchanged as new devices come online.
A sensor registered once is a peer of every existing band from its first observation — its detections place, rank and alert on the same footing.
GSORTS readiness and SensorThings alignment are the named vocabularies a sensor registers against.
Receivers align to established open decoding standards, so commodity hardware integrates cleanly.
RF direction-finding, acoustic and seismic
Beyond the overhead and tracking bands, the covert field carries its own passive modalities. RF direction-finding takes a bearing on an emitter from each receiver; crossed bearings from two or more sites resolve a geolocation fix without ever transmitting. Acoustic and seismic arrays add range and signature — a gunshot is a time-difference-of-arrival across microphones, a footfall is a waveform on a buried geophone. Each is a distinct physical band, classified at the sensor itself, so the field is sensed by emission, sound and ground motion — not only by what an aperture can see.
An emitter that never shows on imagery is still found — its own transmission is the signature.
A third DF site tightens the error ellipse, and every further receiver shrinks it — more geometry, better fix.
Seismic arrays separate footfall, vehicle and digging signatures — ground truth under canopy and in full darkness.
Every passive modality folds into the same shared picture as the overhead and tracking bands, so a fix is one detection among many, not a side channel.
Bring any sensor under one picture.
Talk to us about fielding your bands — a multi-band fusion walkthrough, a covert field-sensor pilot over your AO, or an adapter integration into the picture you already run.