R&D prototype · Passive detection · TRL 4, self-assessed
An RGB camera stream goes to edge AI detection and tracking, then to an alert a person can review. The prototype is built to count false alarms per hour, time to first alert, and track continuity.
An RGB camera stream goes to edge detection and tracking, then to an alert a person can review.
RGB Camera / Video Stream
AI Detection & Tracking
Alert Generation & Human Review
RGB YOLOX-S v3 · Public Anti-UAV dataset · Initial evaluation
99%Event Detection Rate101 of 102 evaluated events
2.53False Alerts / Hour3 false alerts over ~1.19 hours
Public Anti-UAV RGB evaluation: 91 sequences, 102 labeled drone events, and approximately 1.19 hours of video. Three false alerts were observed. The limited test duration makes the hourly false-alert estimate preliminary. Results do not establish performance in real-world deployments.
Thermal, acoustic, mmWave radar, passive RF, and sensor fusion are separate research activities, not part of this RGB path.
View research statusA score on a public dataset does not say how a detector behaves at a port or industrial site. A controlled test has to measure reliability, false alerts per hour, and whether the setup can actually be run there.
A detector can mark a drone in a curated clip and still lose it when the target is small, distant, or partly hidden by weather, glare, or clutter. Field validation checks whether a track holds long enough to raise a useful alert at that site.
Birds, aircraft, insects, wires, and site clutter can look like a drone. On a site, the count that matters is false alerts per hour, and whether a person can review them without the queue becoming the work.
A site test has to fit the camera, power, sightlines, and the people who will review alerts. Placement, hours of operation, privacy limits, and review time are part of the measurement.
The research is aimed at sites that need distributed coverage and a lower operator workload. These are the environments where false alarms per hour would be measured.
Camera coverage of terminals, cargo areas, and waterfronts, where birds, cranes, and vessel traffic are the clutter an operator has to ignore.
Passive awareness along a perimeter, where aircraft and birds are the hard negatives and any test follows site airspace rules.
Detection around plants, substations, and corridors, where a small drone has to be separated from birds and infrastructure.
Sensor placement around yards and plants, where wires, vehicles, and clutter drive false alerts.
A portable RGB setup for a temporary site, with a person reviewing each alert.
Passive detection with an audit trail of alerts and a person in the review loop.
A compact RGB-first setup for sites with limited power, backhaul, or specialist staff.
mAP alone is insufficient for operational systems. A detector can score well on curated frames while still generating hundreds of false alarms per hour in continuous video. Each sensor configuration is scored on the metrics below.
False alarms per hour
Primary metric
First-alert time (seconds)
Also tracked
Track continuity (%)
Also tracked
Cost per km² coverage
Decision metric
Review minutes per shift
Operator metric
Vigiloft is a research and development platform from Galadriel Tech Solutions Inc. for low-cost passive drone detection. The current system is an R&D prototype focused on RGB edge detection, tracking, and measuring false alarms. It is not a production counter-UAS product, and it is not a deployed five-sensor system.
The running path takes an RGB camera stream, detects and tracks with edge AI, and raises an alert for a person to review. Thermal has an offline RGB-thermal fusion benchmark on public data. Acoustic and radar have research baselines. Passive RF exists as a separate Remote ID prototype and is not fused into the RGB path. Module status is listed in the technology section.
ISED treats radiofrequency counter-drone devices that interfere with a control link, or send signals the operator did not intend, as jammers. The Radiocommunication Act generally prohibits jammers. The Minister of Industry can exempt a person, class of persons, or entity by Ministerial Order under subsection 14(1), for purposes such as national security, public safety, national defence, or protection of property, and only on the conditions in that order. A supplier may be covered when working under contract for an already exempted entity, with that use supervised. Those orders are specific. They are not a rule that every legitimate Canadian product must be passive. Vigiloft does not jam, spoof, take over, or kinetically defeat a drone. This prototype detects, tracks, and raises an alert for a person to review.
Birds, aircraft, insects, clouds, wires, and site clutter can resemble a drone to a single sensor. Vigiloft treats false alarms per hour as a primary research metric. Temporal persistence and tracking run in the prototype. Cross-modal confirmation and contextual rules are the next layers in that work.
The project self-assesses at technology readiness level 4. RGB detection runs on the edge, and an early outdoor test has started at a single site with our own 4K camera. Representative Canadian field validation is still required before any operational performance claim.
Ports and maritime facilities, airports, energy sites, industrial facilities, public events, government sites, and remote locations are the environments this research is aimed at. We are looking for organizations willing to host a controlled field test. There are no deployed customer installations.
A first conversation is about a controlled test of the RGB detection path. This is what each side can bring to that discussion.
Galadriel provides
A potential partner participates
Galadriel provides
RGB detection and tracking prototype
A potential partner participates
Coordinate a suitable test site
Galadriel provides
Test plan and performance evaluation method
A potential partner participates
Provide the site environment and operating requirements
Galadriel provides
Detection, false-alarm, and alert analysis
A potential partner participates
Help define feasible controlled test conditions
Galadriel provides
Test results and an improvement report
A potential partner participates
Give feedback on the test results
This is a suggested scope for discussion. It is not a request that you commit to a free deployment, or that you provide site access and support at no cost. It is also not an offer that Galadriel will deploy the system or provide the full service for free. Scope and cost are agreed before any test.
Tell us about the site, the clutter, and what an operator would need to trust an alert.
Or email [email protected].