Detection technology

Why there is no single sensor

Asking which drone sensor is the best is the wrong question. There are four methods, and each of them measures something completely different: a sound, an object, a radio signal or an image. Each has a hard limit beyond which it sees nothing. And the limit of one is typically the strength of another.

A site that operates only one method does not have a detection system, it has a single alarm device with known blind spots.

Acoustics — captures the sound ▾
Rotor blades produce a characteristic frequency pattern made up of the blade passage frequency and its multiples, which can be distinguished from ambient noise. Detection is achieved through microphone arrays with pattern recognition. STRENGTHS: inexpensive, entirely passive, legally uncritical. Covers the area close to the ground where radar sees little. Because of the low price per sensor, acoustics is the only method that allows genuine network density across an area. LIMITS: loss of range in wind, rain and industrial noise. Considerably weaker against quiet electric quadcopters than against large devices with combustion engines. NOT POSSIBLE: identifying, locating the pilot, determining the altitude.
Radar — captures the object ▾
A transmitter radiates and the reflection is evaluated by travel time and frequency shift. This yields position, altitude and speed. STRENGTHS: radar measures the object and not the signal. It is therefore the only method that picks up an autonomously flying, radio-silent drone at all — precisely the case that characterises a prepared attack. LIMITS: bird echoes closely resemble drone echoes. Aircraft, vehicles and weather create background clutter. Small multirotors have a very small radar cross-section. Radar therefore carries the highest false-alarm load. On top of that, radar transmits actively and needs a frequency assignment. NOT POSSIBLE: saying what is flying there and why.
Passive radio detection — captures the radio signal ▾
RF stands for radio frequency, that is, simply radio. RF detection is passive and permitted, whereas RF jamming is an active effect and is not permitted for private parties. What is measured is the entire radio traffic between pilot and device. The downlink carrying video and telemetry is often the stronger source. STRENGTHS: where the protocols can be decoded, the method delivers the serial number, the take-off point and the pilot's location. It is the only method that leads to the person behind the device. Many systems detect a device even before it takes off. LIMITS: blind against radio-silent and autonomously flying systems, blind where a mobile network connection is used. It fails precisely where radar provides cover. NOT POSSIBLE: seeing an object — it sees a signal.
Optics and thermal imaging — delivers the picture ▾
Usually built in two stages, with a wide-angle search and a zoom stage for identification. STRENGTHS: the only solid identification of the object — airframe type, size, attached loads, payload. The only method that delivers material that can later be used in proceedings. LIMITS: needs a clear line of sight and a cue from another method, because a camera cannot lay a search grid across the entire sky. Fog, rain, backlight and darkness impose further limits, although thermal imaging compensates for some of them. NOT POSSIBLE: finding anything on its own. Optics always comes last in the chain.

The layers compared

Four detection layers, compared by what they actually achieve.

↔ Table scrolls sideways

Criterion Acoustics Radar Radio Optics/thermal
measures sound object signal image
active/passive passive active passive passive
detects an autonomous drone partly yes no once cued
leads to the pilot no no yes by chance
identifies the object no no partly yes
in fog yes yes yes no
effort for area coverage low high medium high

The observation that matters: no column says yes all the way down. That is exactly the reason for the fusion layer.

What follows from this:

Network density instead of individual range ▾
The price per sensor decides how many locations can be covered at all — not the peak performance of the individual device. A high-end sensor in one spot covers an area less well than several simple ones in several spots.
A shared situational picture with a confirmation rule ▾
Escalate only once at least two independent methods confirm the same object. Without this rule the false-alarm load becomes unbearable in day-to-day operation — four sensors side by side produce four separate alarm sources and quadruple the false alarms. It is the correlation that turns them into a system.
Cooperative filter ▾
Registered devices from class C1 upwards transmit an electronic remote identification with the registration number, position, altitude and speed, but without personal data. A filter screens out legal traffic before it enters the chain as an event. German airspace is full of legal drone traffic — a system without a filter mainly reports neighbours, surveyors and roofers. Important: the filter does not identify the attacker. It keeps clear the channel in which the attacker remains visible.

Why detection systems fail

They rarely fail because of the technology. They fail because of the false-alarm load. The sequence is often identical:

The system is installed and reports twenty events in the first week. Nineteen of them are birds, neighbouring drones or interference. After three weeks nobody checks the reports any more. After two months the audible signalling is switched off. After six months an expensive system is running that nobody reads.

This is not a technical cause of failure but an organisational one. What helps against it: the confirmation rule, the cooperative filter, a deliberately chosen reporting threshold and a decision on who actually handles a report and when. That last decision is the most important and the cheapest.

The sensor you already have

Before any system is procured, every site already has an observation network covering the whole area and staffed around the clock: the gatehouse, site security, shift personnel, warehouse staff, drivers and maintenance personnel.

What a human being can do

Assess context: a sensor reports an object. A human being has to make sense of the report.

Recognise patterns over time: the same drone, the third evening in a row, the same time, the same approach direction. No single sensor recognises that. This repetition is the single most valuable piece of information there is, because it rules out coincidence.

See the pilot on the ground: a vehicle parked on a farm track, a person with a remote control, a person conspicuously looking up at the sky.

The first thing that needs doing is concrete: this is not primarily about procuring technology, it is about observation skills, a reporting scheme and a named recipient. Those three things cost nothing and take effect immediately.

The reporting scheme

Six fields, deliberately short enough for a notice on the company board:

  1. Time: date, the most precise time possible and the duration of the observation
  2. Location of the observer and viewing direction
  3. Approach and departure direction of the device
  4. Estimated altitude and size, with a reference point
  5. Behaviour: circling, hovering, flying straight through, tracking along a line
  6. Person on the ground, if visible, and vehicles in unusual positions

In addition, where it can be done safely: a photo or video. Your own safety comes first — nothing is touched and nothing is pursued.

What to look out for when observing:

Behaviour
The most telling point. Flying straight through = usually transit. Circling, hovering, tracking along a line at the fence, flying slowly at low altitude = suggests recording. Turning away abruptly on approach = the pilot has line of sight and is nearby.
Direction
Where the device came from and where it disappeared to. Taken together, the two narrow down the pilot's location, because the range within visual line of sight is limited.
Altitude and size
Only a rough estimate is possible — that is fine. Reference points help: level with the hall roof, above the treetops, clearly higher than that. Size in relation to the palm of your hand at arm's length.
Sound
Present or not. A device you cannot hear at low altitude is a different device from one you can hear from far away.
Repetition
Whether the same observation has been made before, by whom and when. This presupposes that earlier observations were recorded. Keeping the records matters as much as making the report.
What does NOT belong in the report
The presumed make or model. Without a zoom shot this is not reliable and creates false certainty during analysis.

Finally

 What typically goes unnoticed:

  • a small device in darkness with no lighting
  • a device above roughly 100 metres while production is running — neither audible nor reliably visible
  • a single straight overflight with no loitering
  • a device over an unobserved part of the site

This is not an argument against vigilance. It is the reason for combining human observation with technical detection, and for the value of every single report.