Contact Us ↗The CIDS carriage monitoring system analyses existing CCTV onboard to detect passenger and crowd incidents, check terminal clearance and flag unsafe driver actions. It retains the existing network, delivers end-to-end alerts within 5 s under test conditions and preserves evidence for review.
Cameras already provide pictures. What is missing is timely interpretation.
Crowding and limited space obscure a single viewpoint, particularly in corners.
Four cameras per car provide a 360° view; a six-car formation supplies 24 feeds.
A train patrol takes 2–5 minutes; incidents can develop before staff arrive.
End-to-end alert ≤5 s; crowd-event response ≤3 s.
Crew report via a supervisor and the cab before information reaches the centre.
Structured alerts go directly to the cab and control centre and are synchronised to the cloud.
Fatigue and attention affect the consistency of around-the-clock patrols.
Continuous monitoring and reviewed examples support model refinement. Crew remain responsible for intervention.
Four dome or fisheye cameras per car provide 360° coverage and 1080p RTSP feeds.
AFR600 performs onboard inference and buffers data in terabyte-class local storage during outages.
PIS switches and central servers handle timing, commands, device management, data cleansing and distribution by priority.
Cab pop-ups, live views, centre alerts and cloud records support tiered notification.
RTSP capture → PIS network → AFR600 inference → event record, marked image and preceding 30 s of video → central archive and alert display
Take RTSP video from the PIS switch without changing cameras or wiring.
Dedicated nodes operate without consuming the CCTV host’s computing capacity.
The centre synchronises clocks and starts or stops detection remotely.
One node serves three or four cars. In this six-car example, cars 03 and 04 each carry an AFR600.
Synchronised timestamps · Health checks every minute · Video ≥7 days, events ≥14 days, logs ≥30 days · Two-way PIS communication and resumable uploads
Figures refer to standard test conditions. Results depend on camera installation, light, crowding and software version; agreed project tests govern acceptance.
Onboard inference identifies movements and event features in the video.
Classify the event, locate the carriage, establish the time and filter interference.
Mark the image and save a 30-second recording around the incident.
Send encrypted reports over the dedicated PIS network to central and cloud services.
Display a cab alert, monitor at dispatch and retrieve the historical record.
Average response <2 s · event retention 100%
These are workflow test results, not per-function response limits. Recording windows are project-specific.
Cycles through four views and opens live video with event type and location on alarm.
Receives train health and events and issues clearance checks, clock updates and emergency dispatch instructions.
Monitor multiple trains, roll out remote updates, stream live video and review history.
Complementary views reduce single-camera occlusion. Test detection accuracy exceeded 98%.
Track 15 body joints over time to distinguish walking, falls, reclining and waving for help.
Match pedestrian features and tracks across views to associate observations of one person; camera-level reports remain independent.
Analyse onboard and transmit key events rather than all feeds. In testing, recognition exceeded 98% and processing latency was below 200 ms.
Figures refer to standard test conditions. Results depend on camera installation, light, crowding and software version; agreed project tests govern acceptance.
| Dimension | CIDS | Crew patrols |
|---|---|---|
| Detection and response | Continuous monitoring with alerts in seconds | Patrol cycle of 2–5 minutes |
| Peak-hour visibility | Continuous coverage from 24 feeds | Fixed views obstructed by crowds |
| Continuity | Designed for round-the-clock operation | Affected by fatigue and attention |
| Accumulated experience | Models refined with accumulated data | Staff turnover and repeated training |
| Incident response | Automatic alerts with event evidence | Staff can intervene and help on the spot |
| Service consistency | Consistent configured criteria | Varies with individual experience |
CIDS monitors and records evidence. Crew assess the scene and can provide immediate assistance.
In-service monitoring · Terminal clearance
Deployment on a six-car metro train covered installation, configuration, centre integration and acceptance. The line name and site footage are withheld at the customer's request.
Last updated