LiDAR-based track monitoring detects obstructions, trespass and infrastructure hazards in darkness, glare and weather that defeat cameras. The engineering difficulty is not detection — it is discriminating a person or a rockfall from a fox, a plastic bag or heavy rain, reliably enough that the railway keeps the system switched on.
Range comes directly from LiDAR rather than being inferred, which is why it holds up where a camera's assumptions do not.
| Signal | How | What it changes |
|---|---|---|
| Obstruction on the track | Point-cloud deviation from the learned clear envelope | A vehicle, debris or fallen load in the running path, detected in darkness |
| Trespass and intrusion | Object classification within a defined trackside zone | People on or approaching the track, at stations, crossings and known access points |
| Level crossing occupancy | Volumetric monitoring of the crossing box | A vehicle stopped or trapped on the crossing as a train approaches |
| Rockfall and slope movement | Comparison against a reference surface over time | Cutting and embankment material arriving on the track, and slow slope creep before it does |
| Structure gauge intrusion | Envelope violation against the permitted gauge | Vegetation, scaffolding or plant fouling the space a train needs |
| Platform edge | Presence detection in the gap and on the edge line | A person or object in the danger zone as a service arrives |
The sensing is the solved part. These are what determine whether the deployment is still running in year three.
A stopped service costs delay minutes, compensation and public confidence. A system that stops trains for a fox will be overridden within weeks, and an overridden system protects nobody — so precision is weighted as heavily as sensitivity, which is the opposite of most detection work.
Rain, fog and snow return energy that looks like an object. Discriminating a genuine obstruction from precipitation backscatter requires multi-echo handling and temporal consistency, and the system must degrade its own confidence honestly rather than keep asserting.
The classes that matter are defined by consequence, not by appearance. A fox, a bin bag and a child produce comparable returns at range, and separating them is where the engineering effort actually goes.
Mounting on trackside structures subject to vibration, thermal movement and passing trains means extrinsic calibration shifts over months. Continuous self-monitoring against fixed reference geometry is required, because a silently misaligned sensor reports a clear track.
Many of the locations that most need monitoring have neither. Power budget, local processing and store-and-forward are architectural decisions, and sending point clouds anywhere is not an option.
Anything that can influence train movement falls under rail safety standards, with an independent argument and evidence. Most deployments are therefore advisory to a control room first, and the integration path is agreed with the infrastructure manager before design.
Corridor, crossing, station and depot.
Obstruction and trespass detection across sections with a history of incursion, rockfall or fly-tipping.
Volumetric detection of vehicles trapped or stopped on the crossing, integrated with existing crossing control.
Presence detection in the platform-train gap and on the edge line as services arrive.
Personnel and plant proximity in shunting and maintenance areas where movements are frequent and sightlines are poor.
LiDAR measures range directly rather than inferring it, and works in darkness, glare and high contrast where a camera's assumptions break down. In practice the two are usually fused — LiDAR for geometry and presence, camera for classification and evidence.
Not detection. Discrimination — separating a person or a fallen load from a fox, litter or precipitation backscatter, reliably enough that the railway keeps the system enabled. A system overridden by staff protects nobody, whatever its detection rate.
Rain, fog and snow return energy that resembles an object. Multi-echo handling and temporal consistency separate genuine obstructions from precipitation, and the system reports reduced confidence honestly rather than continuing to assert in conditions it cannot see through.
Anything influencing train movement falls under rail safety standards and requires an independent safety argument with supporting evidence. Most deployments are advisory to a control room first, with the integration path agreed with the infrastructure manager before design begins.
Local processing with a power budget designed for it, and store-and-forward for events rather than raw data. Transmitting point clouds off site is not viable, so the classification decision is made at the sensor.
Rail and vehicle sensing systems.
INDUSTRYWhy false alarms decide adoption.
AILiDAR, camera and radar aligned in time.
TECHNOLOGYThe safety argument and its evidence.
AIZone intrusion and classification.
SOLUTIONDistributed trackside assets.
Tell us the specification, the constraint and the deadline. Programmes that cross silicon, radio, embedded and AI are where Faststream is strongest.