Intelligent Video Analytics Software for Intelligent Video Surveillance and Monitoring
Surveillant is intelligent video analytics software that runs deep-learning models against the cameras you already own. It classifies people, vehicles, and behavior in every frame, understands the context of a scene rather than counting moved pixels, alerts only on what you defined as important, and lets an operator find any moment by describing it in plain English.
Runs on any ONVIF or RTSP camera, NVR, or existing VMS. No new hardware required.
- Also called
- IVA, video intelligence
- Detection method
- Deep-learning neural networks
- Where it runs
- Cloud, on your streams
- Search
- Natural language
- Best for
- US multi-camera operators
What Is Intelligent Video Analytics?
Intelligent video analytics (IVA) is software that uses computer vision and deep-learning neural networks to interpret video footage automatically. It detects and classifies objects, tracks them across a scene, recognizes behaviors and anomalies, and generates searchable metadata about what happened, so a surveillance system reports events instead of only recording them.
The word that matters is intelligent. Standard video analytics applies fixed rules to pixel changes: motion in a box, a line crossed, an object left behind. Intelligent video analytics applies trained models that recognize what an object actually is and what it is doing, which is why it can tell a delivery driver from an intruder, a shopping cart from a person, and a shadow from a car.
In practice IVA is now a software layer, not a camera feature. Surveillant reads streams from the ONVIF and RTSP cameras, recorders, and video management systems you already run, and applies detection, anomaly alerting, and natural-language search in the cloud. Nothing is ripped out, and cameras bought years ago get the same models as cameras bought this quarter.
IVA does four things a recorder cannot:
- Classifies what an object is, not that it moved
- Reads behavior and context inside a scene
- Learns what is normal and flags anomalies
- Writes metadata that makes footage searchable
Intelligent Video Analytics vs Standard Video Analytics
Both are called video analytics on a spec sheet. Only one of them knows what it is looking at. This is the line buyers most often fail to check before signing.
| Dimension | Standard video analytics | Intelligent video analytics (IVA) |
|---|---|---|
| Underlying method | Pixel-change rules and thresholds | Deep-learning neural networks |
| What it detects | Motion, line crossing, object left behind | Object class, identity attributes, behavior, anomaly |
| Scene context | None, every trigger looks alike | Understands zone, time, object type together |
| Weather and lighting | Rain, headlights, shadows all trigger alerts | Filtered out because they are not classified objects |
| Improves over time | No, rules are static until retuned by hand | Yes, models are retrained and redeployed |
| Searching footage | Scrub by camera and timestamp | Query by description across every camera |
| Typical false alerts | High enough that teams mute the channel | Low enough that an alert queue gets worked |
| Where it runs | On the camera or recorder, per device | On the edge or in the cloud, across all cameras |
Newer to the topic? See the plain-language explainer on video analytics vs CCTV, or read how the models work in how AI video analytics works.
The Four Generations of Video Analytics
Vendors use "intelligent video analysis" to describe all four. Knowing which generation you are being sold is the fastest way to judge a quote.
| Generation | How it decides | Question it can answer |
|---|---|---|
| Video motion detection | Counts changed pixels against a threshold | Did anything move? |
| Rule-based analytics | Geometry rules on a moving blob | Did a blob cross this line? |
| Deep-learning IVA | Trained neural networks classify each object | Was that a person, a truck, or a raccoon? |
| Vision-language search | Models that map video and text into one space | Show me a white SUV in the north lot after 9pm |
Surveillant runs the third and fourth generations together: classification for alerting, vision-language search for investigation. Compare the categories side by side in our breakdown of the types of video analytics, then work through the AI video analytics implementation plan before you connect the first camera.
What Intelligent Video Analytics Software Delivers
Every capability below runs against streams you already have, from the first camera you connect.
Object Detection and Classification
People, vehicles, bicycles, packages, and weapons are identified by class in every frame, so a rule can fire on a person in a restricted yard and stay silent for a passing truck.
Behavior and Anomaly Analysis
Loitering, tailgating, crowding, fighting, slip and fall, and wrong-way movement are recognized as patterns over time rather than as single-frame events.
Natural-Language Video Search
Describe what you are looking for and get matching clips across every camera and every day of retention. This is the capability that turns an archive into an investigation tool.
Cross-Camera Object Tracking
Follow one person or vehicle as it moves between camera views and assemble a single movement timeline for a site instead of a folder of disconnected clips.
Real-Time Intelligent Monitoring
Intelligent video monitoring runs continuously on every feed at once. Software attention does not drift after twenty minutes the way human attention does.
Searchable Scene Metadata
Each detection is written as structured metadata: object class, attributes, zone, timestamp, camera. That metadata is what makes retrieval fast and what feeds your reporting and API.
How Intelligent Video Analytics Works on Your Cameras
Four steps, usually finished inside a week for a single site. Your existing recording never stops.
Ingest the Streams
Point Surveillant at RTSP URLs, an ONVIF discovery range, or your current NVR or VMS. Frames are decoded and queued for inference. Nothing on your side is reconfigured.
Detect and Classify
Neural networks label every object in every frame, then a tracker links those labels across frames so one car crossing a lot is a single tracked object, not eighty detections.
Apply Context and Rules
Object class, zone, dwell time, direction, and hour combine into the rule that decides whether anything is worth an alert. A person on a dock at 2am is not a person on a dock at noon.
Alert, Search, Integrate
Classified events reach the dashboard, mobile, email, or a webhook, and every detection lands in the search index and the API for your own reporting.
Intelligent Video Analytics Solutions by Environment
The teams that see the fastest payback share one trait: more cameras than people who can watch them.
Multi-Site Retail
Sweep thefts, register incidents, and after-hours entry surfaced across dozens of stores from one console, with queue and dwell metrics coming off the same feeds.
Warehouses and Logistics
Dock door activity, trailer theft, and unsafe movement around forklifts, with anomaly detection covering the shifts when nobody is on the floor.
Manufacturing and Industrial
PPE compliance, restricted-zone entry, and line stoppage all read from cameras that already exist for security, so one deployment serves safety and operations.
Critical Infrastructure
Fence-line intrusion at unmanned sites where every false alarm means dispatching a truck. Object-class rules are what keep response cost down.
Schools and Campuses
Weapon detection and perimeter monitoring across buildings, with alerts that reach staff and responders in seconds rather than after a phone tree.
Property and Facilities
Lobby tailgating, garage incidents, and tenant disputes reduced to a two-minute natural-language lookup instead of an afternoon of scrubbing.
Intelligent Video Analytics Questions
What is intelligent video analytics?
Intelligent video analytics (IVA) is software that uses computer vision and deep-learning neural networks to interpret video footage automatically. It detects and classifies objects, tracks them across a scene, recognizes behaviors and anomalies, and produces searchable metadata about what happened, so a surveillance system reports events instead of only recording them.
What is the difference between intelligent video analytics and standard video analytics?
Standard video analytics applies fixed rules to pixel changes, so it knows something moved but not what it was. Intelligent video analytics uses trained neural networks that classify the object and read its behavior in context. That is why IVA can separate a person from a shadow, a truck from a raccoon, and an intrusion from rain on the lens.
What does IVA stand for in video surveillance?
IVA stands for intelligent video analytics. It refers to the AI software layer that analyzes surveillance video to detect objects, behaviors, and anomalies in real time. Some vendors use IVA as a product name for their on-camera analytics package, so always confirm whether a quoted IVA feature runs deep-learning models or older rule-based detection.
Is video analytics considered AI?
Modern video analytics is AI. Deep-learning object detection, behavior recognition, and vision-language search are all machine-learning techniques. Older motion detection and rule-based analytics are not AI: they are threshold and geometry logic. Both ship under the label video analytics, which is why buyers should ask which one a quote actually includes.
How does intelligent video analytics work?
The software ingests camera streams over RTSP or ONVIF, decodes frames, and runs neural networks that label every object. A tracker links those labels across frames into tracked objects. Rules combine object class, zone, dwell time, direction, and hour to decide what deserves an alert, and every detection is indexed so footage becomes searchable.
Can intelligent video analytics work with existing CCTV cameras?
Yes, in most deployments. Software IVA connects to any camera that supports RTSP streaming or the ONVIF protocol, which covers most IP cameras sold in the past decade, and it can also pull streams from an existing NVR or VMS. The limit is image quality: no model recovers detail a sensor never captured at night.
How much does intelligent video analytics software cost?
Analytics-only software generally runs about $3 to $15 per camera per month based on 2026 vendor and reseller estimates, layered onto cameras you already own. Full cloud platforms that bundle AI, storage, and alerting, including Surveillant, typically fall in the $39 to $42 per camera per month range. Bundled proprietary cameras add roughly $600 to $3,500 each up front.
What are the main use cases of intelligent video analytics?
The common ones are perimeter and intrusion detection, weapon detection, loitering and tailgating alerts, PPE and safety compliance, license plate recognition, retail loss prevention, people counting and dwell analysis, and forensic search after an incident. Most operators start with one alerting use case and adopt search once the metadata exists.
Related Reading
AI Video Analytics Software
The full detection platform, capability by capability.
AI Video Management System
Recording, search, and alerting in one AI VMS.
How Accurate Is AI Video Analytics?
What accuracy numbers mean and how to test them.
AI Video Analytics Cost
Pricing models, cost per camera, and payback.
See intelligent video analytics on your own footage
Connect a few streams, draw a zone, and watch classified detections and natural-language search run against video you already record. No credit card required.