The perimeter used to be a civil-engineering product: posts, mesh, razor, done. The current generation is a sensing instrument. Distributed acoustic sensing turns the fence fiber itself into a microphone that classifies a climb versus wind versus a rabbit. Ground radar tracks a walker at hundreds of meters and slews a camera onto them automatically. Lidar draws virtual walls across open ground. Analytics watch every stream so no human has to pretend they can.
Every spec sheet promises detection. The honest ones price the false alarms.
The two numbers that matter
Probability of detection sells systems; nuisance alarm rate decides whether they work. A perimeter that cries wolf forty times a night trains your operations center to ignore it by Thursday, at which point its true probability of detection is zero regardless of the brochure. When we write perimeter specifications, the binding line is the pair: PD above a threshold at a defined intrusion set, NAR below a ceiling in your weather, your wildlife, your traffic. Then we test it on site before acceptance, in the rain, with the trucks running.
The stack that earns its keep
- Detect at the line: DAS or taut-wire on the fence, radar over the open ground, chosen for the terrain rather than the demo.
- Verify by camera: every alarm auto-slews a PTZ so a human sees truth in seconds, not a map dot.
- Delay at the building: the high-tech fence buys minutes only if doors, docks and mantraps spend them.
- Respond from a post that trusts the system, because the false-alarm budget was engineered, not endured.
None of this replaces the older discipline; it instruments it. The fence grew a brain. The acceptance test is where you find out if it grew judgment.
We specify and acceptance-test perimeters in real weather. Your numbers deserve to be measured, not quoted.