Video analytics
Video analytics is what turns a camera stream into an event a person can act on. This page describes the three kinds of detector the platform runs, the analytics suites they are grouped into, and the operational scenarios each one is built for.
The scenarios listed here are the ones the platform executes today. The catalog grows with each release, so check What's new for additions. Every scenario is subject to the same constraints described in Engineered for ethics: recognition applies to pre-registered persons and vehicles, and privacy-sensitive searches require a Case ID.
Three kinds of detector
Modules differ in what they compare against, and that difference determines how they are configured and what they cost to run.
| Kind | How it works | Examples |
|---|---|---|
| Trackers | Detect and follow objects frame by frame at high frame rate, then classify their attributes. No reference list is involved. | ObjectTrack Pro, MotionTrack Pro, CrowdCount, CarCount, GunTrack, FireTrack, TamperTrack |
| Matchers | Compute a descriptor for a face or a plate and compare it against entries pre-registered in a watchlist. A match raises an event; a non-match is not retained as an identity. | FaceTrack Pro, Plate number recognition |
| Prompt-defined detectors | A Vision-Language Model evaluates sampled frames against a condition written in plain text. No dataset, no training, no new module to deploy. | StreamVLM™ (beta, selected instances) |
Real-time and investigative use
Real-time analytics exist to cut incident response time in control rooms, real-time crime centers, and fusion centers running 24/7. An intruder entering a protected area needs an immediate response, so a module handling that class of incident, such as ObjectTrack Pro, has to perform reliably in real time. The same applies to a person on a rail track, a fire, the appearance of a missing person, or a highly wanted suspect. With customizable alarm monitors and the urgent notification system, response time falls to roughly one second in most cases.
The same modules also serve post-incident work: reconstructing a suspect vehicle's route with CarTrack Pro, or identifying possible accomplices with FaceTrack Pro. Here the Searchveillance™ tool finds events of interest across the whole archive against user-defined criteria, and accuracy matters more than latency.
Whether a module counts as real-time or investigative depends on the use case, not on the module. Used to search for a missing child, FaceTrack Pro is a real-time tool. Used in a post-incident investigation, it is an investigative one. That is why these modules combine high accuracy with substantial computational capacity rather than trading one for the other.
The analytics suites
Modules are grouped into suites, each covering one operator-facing capability. Which modules can run together on the same camera channel is documented in Module compatibility; the image quality each one needs is in the Camera requirements guide.
| Suite | What it does | Where it is documented |
|---|---|---|
| Face and facial features | Watchlist suspect recognition; attribute search by gender, age, ethnic group, facial hair, and glasses when no suspect photograph exists. | FaceTrack Pro |
| License plate and vehicle features | Plate recognition plus make, model, color, type, and body style; stolen, uninsured, and unregistered vehicle detection through external database integration. | CarTrack Pro, Plate number recognition, CarCount |
| Perimeter security | Intrusion detection, restricted-zone monitoring, abandoned-object and loitering detection for critical infrastructure, campuses, and transport hubs. | ObjectTrack Pro, MotionTrack Pro |
| Railway and transportation safety | People on tracks, train surfing, platform-edge crowding, and unattended luggage in transit environments. | ObjectTrack Pro, CrowdCount |
| Crowd analytics | Real-time dangerous-crowding alerts and precise post-event counting, up to 3,000 people in a region of interest. | CrowdCount |
| Person re-identification | Appearance-based tracking by clothing texture, accessories, and hat or hair color when the face is occluded, masked, or off-angle. | Find a person |
| Multi-camera traffic analytics | Red-light, stop-line, wrong-way, forbidden-maneuver, illegal-parking, right-of-way, railway-crossing, and average-speed enforcement; two-wheeler detection; congestion analytics. | Manage traffic at crossroads, Bike Lanes |
| Universal VLM analytics | Custom detectors from plain-text prompts, with no training and no dataset collection. Up to 100 prompt-defined detectors per camera channel. Beta, selected instances. | StreamVLM |
| Access-control biometrics | Face-based access control with liveness and anti-spoofing checks at sensitive entry points. | FaceTrack Pro: access control |
| Fire and smoke alarm | Early fire and smoke detection across indoor and outdoor scenes. | FireTrack |
| Camera tamper and quality control | Detects obstruction, defocus, repositioning, and quality degradation, and raises maintenance work tickets automatically. | TamperTrack |
| Audio analytics | Detection of sounds relevant to public safety on channels that carry an audio stream. | AudioTrack |
Popular use cases
Find a suspect
In this use case, a real-time search runs for a person strongly suspected of serious criminal activity, with the goal of preventing further harm. When the suspect's identity or appearance is known, the system detects a facial match in the incoming video stream in real time and alerts the operator. The current location and the previous route are reported as well, so law enforcement can respond immediately.
Find a missing child
The platform integrates with external databases, for example archives of wanted persons, missing children, or sex offender records. Those databases are maintained only by authorized institutions. Contact the Support Team to import such a database through the API. A continuous search then runs across the monitored area, matching incoming video against the database entries.
Reveal a suspect's connections
A post-incident investigation often has to identify everyone involved. The platform identifies a user-defined person from a screenshot, finds a facial match in the available person databases, or reconstructs that person's route within the coverage area.
Find a license plate
The suspect's face may be unknown while the vehicle plate is known. In that case, the "Find license plate in real time" use case raises an alert whenever the vehicle appears.
Find a vehicle by its features
The route of a suspect's car can be strong evidence at trial. The platform finds where and when a vehicle was detected, from its plate number or from visual characteristics such as type and color.
Detect a person on a rail track
Railways and underground stations need strong protection against unauthorized track access. When an intrusion happens, the platform detects it in real time and raises an alarm so the response is immediate.
Find abandoned luggage
Abandoned luggage can be a serious threat to public security, especially in transport hubs and other crowded spaces. A dedicated module detects such objects in real time, which contributes directly to counter-terrorism work.
Detect firearms
GunTrack is trained to detect visible firearms in real time at entrances to areas where guns are prohibited. Because the danger is immediate, incidents of this kind sit at the top of every responder's priority list.
Detect perimeter intrusion
A routine duty of the security service at a restricted facility is to monitor its perimeter and watch for suspicious activity inside. The platform offers several real-time modules for this, configurable for a custom area or line of protection, for closed or open spaces, and for sensitivity to object type.
Other threats to a restricted facility come from suspicious behavior, such as loitering, or from accidental gaps in protection, such as guards missing from their posts. There is a module for each.
Manage biometric access
The platform manages people's access to premises and vehicles' access to a parking lot. It searches for a facial or plate match in the live stream. On a match it opens a smart lock. An unknown person or vehicle at the entrance raises an operator alert instead.
Detect dangerous crowding
Dangerous crowding can build on a railway platform, at an elevator entrance in a transport hub, or in other high-risk areas. The platform counts people within a user-defined area in real time and alerts the operator when the number crosses a set threshold.
Control traffic
The platform supports effective traffic management, for example across an urban district, by reporting congestion, traffic rule violations such as speeding and illegal lane crossing, and illegal parking.
Detect a condition with no dedicated module
Many municipal conditions never justified a purpose-built module. Examples are standing water in an underpass, illegal dumping, fresh graffiti, a damaged fence, uncleared snow on a fire exit, and a person lying on the ground. StreamVLM covers these from a plain-text description, without a dataset and without model training.