Third alert since midnight
The fence camera on the north perimeter trips at 0214. Person-shaped motion near the wire. The operator on duty pulls the feed and watches a deer work the ditch line for the second time tonight. He clears the alert. At 0247 the camera trips again and he clears it without looking. By 0300 the system has trained him to ignore it.
Single-sensor security does not fail by missing things. It fails by detecting everything and explaining nothing. A motion alert is a pixel change. It cannot tell a deer from a scout or a delivery truck from a probe. So it fires on all of them and the operator learns that the alarm means nothing. The night the alert is real looks exactly like the two hundred nights it was not.
At 0322 a pickup comes up the access road with its headlights off. On the single-camera system that is one more motion event in a night full of them. On EdgeLance it is something else entirely.
Six ears on the same fence
EdgeLance runs six sensing modalities on hardware the team already owns. Video runs detection models against any camera feed. It finds people and vehicles, watches zones, matches faces against a mission watchlist and reads license plates. Acoustic listens through any microphone. Drone motors show up as harmonics in the 90 to 420 Hz band. Vehicle engines sit between 40 and 180 Hz. Gunshots have a signature of their own, and voice activity flags a human presence the camera has not seen yet.
RF is the third ear. A software-defined radio dongle picks up drone control links, ADS-B aircraft tracks and emitters that have no business being near the wire. NFC badge presence tells the system which contacts are friendly before anyone keys a radio. And wearable biometrics cover the blue side. The fabric tracks the threat outside the wire and the readiness of the team inside it.
None of this gear is exotic. Cameras and microphones and a radio dongle are commodity hardware. What changed is where the fusion happens. Every modality feeds a correlation engine on the node itself, so the threat picture assembles on the same box that owns the sensors even when the uplink is gone. Six streams become one contact before anything leaves the machine.
The truck told on itself
Back to the pickup at 0322. The camera sees a vehicle running dark. The acoustic sensor confirms an engine idling in the 40 to 180 Hz band. That alone upgrades a motion event to a corroborated contact. But the RF layer is what turns it into a story.
Every modern vehicle constantly broadcasts the unique IDs of its tire pressure sensors. The transmissions are unencrypted and the serial numbers never change for the life of the tire. Almost nobody listens. EdgeLance listens with a $30 software-defined radio dongle. It decodes more than 28 TPMS protocols and logs every tire ID it hears, which makes any vehicle that passes the wire twice a vehicle the system recognizes. Passive re-identification at range, off a broadcast the target does not know it is making.
So the contact card in front of the operator at 0322 does not say motion detected. It says a vehicle running dark with an engine confirmed on acoustic. And it says the tire IDs match a truck logged near the east gate twice last week. One correlated contact with a history instead of three alerts with none. The operator who cleared two deer an hour ago is now looking at something worth waking people up for.
Every detection carries its receipt
Fusion is only trustworthy if you can unwind it. Every detection in the fabric carries a hash link back to its source. The video hit points at the exact frame. The acoustic hit references the recorded segment. The TPMS record links the raw signal capture. When a contact card says a truck has been here before, the operator can pull the original evidence in two clicks and judge it himself.
That matters after the fact as much as during. When the incident goes to review, the board wants to know what the system saw and when it saw it. A fused contact with hash-linked evidence answers from its own record. The alternative is three separate sensor logs with three separate clocks and an analyst reconciling timestamps by hand. EdgeLance writes the correlation into the mission record at detection time, signed and attached to the frames and signals that produced it.
The fence line already needed this
US bases are living this problem now. NORTHCOM has tracked hundreds of drone incursions over domestic installations, and by its own accounting it defeats about one in four of the drones it detects. Part of that gap is authorities. Part of it is sensing. A drone that shows up on one modality is a blip a defender hesitates on. A drone confirmed on camera, on motor harmonics in the 90 to 420 Hz band and on its RF control link is a track someone can act against.
Ukraine reached the same conclusion under harder pressure. CSIS documents how its military went digital from the bottom up. Units stitched commodity sensors into a common picture because the integrated systems they were promised never arrived in time. Fusion turned out to be a software problem, and software can run at the node on whatever the team already carries.
The sensor fabric ships as part of EdgeLance on Windows, Android, macOS, iOS and Linux, and it treats anything the team plugs in as another ear on the fence. The pickup from 0322 is in the mission record now. Tire IDs, engine band, dark approach and both prior sightings. The next time those tires broadcast near the wire, the contact card opens with the full history already attached.