Video surveillance explained for business buyers: how a video surveillance system works, its parts, AI analytics, deployment choices, privacy law and costs.
Video surveillance is cameras plus software that watch, record and analyse activity to protect people and property.
The camera is the smallest decision. Recording, software, analytics and monitoring decide whether the system actually works.
Deployment shapes everything: on premise, cloud, hybrid, edge or air gapped each trade control against convenience.
AI video surveillance earns its keep when it decides which events matter and acts on them, not when it adds more alerts.
Privacy law applies the moment a camera sees people. Signage, purpose and retention periods are not optional.
Licensing models differ more than prices. Know what you pay per camera, per year and for AI before you sign.
Plan from the risk, not the catalogue: decide what you must see and do, then choose the equipment.
Search for video surveillance and most results fall into two camps. One sells cameras for the home. The other offers a short definition and a list of benefits. Neither helps a security head, a facilities manager or a CIO who has to choose, defend and run a system for years.
What gets left out is how the parts fit together, what each deployment model costs you in control, what AI can and cannot do, and where the law draws lines. This guide covers each topic at decision depth and points to our dedicated guides for more.
Video surveillance is the use of cameras and software to watch, record and analyse activity in a place in order to protect people, property and operations. A video surveillance system captures video, stores it, shows it to operators live and on playback, and increasingly uses AI analytics to detect events and trigger a response automatically.
People also call it CCTV, video security or video monitoring. CCTV, short for closed circuit television, is the oldest name, from a time when cameras fed a private set of screens rather than a broadcast. The terms now overlap almost completely.
Closed circuit television first appeared in the 1940s, famously in Germany in 1942 to watch rocket launches. Video cassette recorders in the 1970s made recording practical, and digital multiplexing in the 1990s let one recorder capture many cameras. Axis Communications released the first network camera in 1996, and IP cameras gradually replaced analogue ones over the following two decades. The current shift is from recording to understanding: AI analytics that recognise people, vehicles and events, and software that acts on them without waiting for someone to notice.
Every video surveillance system, from a single shop to a city, is built from the same eight parts.
Cameras. The eyes of the system: bullets and domes, PTZ cameras that pan, tilt and zoom, fisheye, thermal and plate reading cameras. Our guide to CCTV camera types compares them, and our IP camera complete guide explains resolution and lenses.
Network and PoE. IP cameras send video over an ordinary Ethernet network. Power over Ethernet (PoE), defined in the IEEE 802.3 standards, carries power and data on a single cable, which simplifies installation. Our guide to Power over Ethernet covers switch budgets and cable runs.
Recording: DVR, NVR and VMS. A DVR records analogue cameras wired over coaxial cable. An NVR records IP cameras over the network. A VMS records in software on servers you choose, across any number of sites. Our NVR vs DVR guide goes deeper.
Storage. Disks or object storage hold the footage. How much you need depends on camera count, resolution, frame rate, codec, recording mode and how many days you keep. Efficient codecs such as H.265 store more days on the same disks.
Video management software. The VMS is the brain. It connects cameras of different brands, manages recording and retention, shows live and recorded video, controls users and permissions, and links to other systems. ONVIF, the industry standard that lets cameras and software from different makers work together, is what makes a mixed estate manageable.
Analytics. Software that interprets video: detecting intrusions, reading number plates, counting people, spotting fire or missing safety gear.
Monitoring and the control room. Operators watch live video on workstations and video walls, respond to alarms and coordinate a response. Our video wall guide covers control room design.
Evidence and export. When something happens, footage must be found fast, exported, and proven untouched. Good systems lock evidence, seal it so tampering is detectable, and redact bystanders before sharing.
A video surveillance system works as a chain: capture, transport, record, analyse, alert, respond and preserve. Each link can fail on its own, so judge the whole chain.
1. Capture. The camera turns light into a compressed video stream, usually H.264 or H.265, often a main stream for quality and a sub stream for lighter viewing.
2. Transport. The stream crosses the network to the recorder or VMS, typically over RTSP, the long established protocol for camera video.
3. Record. The VMS writes video to storage continuously, on a schedule, or when an event fires, and deletes it when the retention period ends.
4. Analyse. Analytics inspect the video for events that matter: a person in a closed zone, a vehicle on a watchlist, smoke, a fall.
5. Alert and act. Rules decide what happens next: notify a team, raise an alarm in the console, sound a siren, speak a warning through the camera, or message another system.
6. Respond. An operator views the live feed and coordinates people on the ground.
7. Preserve. The relevant footage is locked and exported as evidence.
For most purposes there is no difference: the two terms describe the same hardware. Where people draw a line, it is about the system around the camera rather than the camera itself.
A security camera is often a standalone device protecting one home or shop. A surveillance camera is part of a managed system with central recording, operators, analytics and an audit trail. A business needs the second, whatever the box on the wall is called.
Where the software runs and where the video lives is the most consequential choice you make. Each model has honest trade offs.
On premise. Servers on your site run the VMS, recording and analytics. You control the data, the network and the upgrade timing, and live view stays fast because nothing leaves the building. You also own the hardware and its upkeep.
Cloud and VSaaS. Video surveillance as a service runs the software in the vendor's cloud, and often records there too. Setup is quick and there is little hardware to manage. In return, you pay a recurring fee per camera, depend on internet bandwidth, and hand your video to a third party. Our VSaaS guide covers when it fits.
Hybrid. Recording and analytics stay on site while archive or remote access uses the cloud. It balances control and convenience, with clear rules about what leaves the site.
Edge. Recording or analytics run close to the cameras, in the camera itself or a small server at each site. It saves bandwidth and keeps working when the link to head office fails. Our edge AI video surveillance guide explains the design.
Air gapped. The system has no internet connection at all, the norm for defence, critical infrastructure and some government sites. It demands software that licenses and updates offline. Our air gapped deployment guide walks through it, and our cloud vs on premise comparison weighs the five year costs.
AI video surveillance uses software models to recognise what is in the video, people, vehicles, objects and events, so that the system can decide what matters and act without waiting for an operator to notice. Done well, it watches, decides and acts.
Watch. Analytics detect things and behaviours: intrusion into a zone, line crossing, loitering, a number plate on a watchlist, a crowd forming, a fall, fire and smoke, a missing helmet, an abandoned bag, a covered or turned camera.
Decide. Rules combine detections with context. A person at the loading bay at two in the afternoon is normal. The same person at two in the morning, after a door forced alarm, is an incident. Good rules cut alerts rather than adding them.
Act. The system notifies the right team, raises an alarm, triggers a siren or relay, speaks a warning through the camera, or sends the event to a security operations platform.
Be clear about limits. AI depends on the picture: placement, light and resolution still matter. Face recognition carries the heaviest legal duties of any analytic. Our AI video analytics enterprise guide covers the analytics in depth, and our ethical AI surveillance guide covers responsible use.
Score every shortlisted system against the same ten questions.
1. Camera compatibility. Does it work with the cameras and recorders you already own, across brands, through ONVIF and RTSP? Replacing working cameras is the most expensive way to modernise.
2. Scale. Can it grow from one site to many, with central management and a standby server for continuity?
3. Latency. How far behind real life is live view? For an operator directing a guard, a delay of several seconds matters. Ask how live view reaches the browser.
4. Analytics. Which analytics are built in, how are they licensed, and can zones and sensitivity be set per camera?
5. Automation. Can rules combine conditions, sequences and schedules, and trigger real actions such as relays, talkdown and integrations?
6. Evidence integrity. Can exported footage be proven untouched, and can bystanders be redacted before sharing?
7. Security and access control. Does it support single sign on, granular permissions and a tamper evident audit trail? Does it connect to your door access system?
8. Privacy. Are there tools for retention limits, redaction, and consent and erasure for any biometric watchlists?
9. Deployment. Can it run where your policy requires: on premise, edge, hybrid or fully offline?
10. Total cost and licensing model. What will you pay in year one and year five, per camera, per site and for AI? Our VMS pricing comparison and our comparison of the best AI surveillance companies show how vendors differ.
Video of identifiable people is personal data in most of the world, so a video surveillance system carries legal duties from the first day. Laws differ, and this is general guidance, not legal advice.
In Europe, the GDPR applies, and the European Data Protection Board's guidelines on video devices set out what a lawful system looks like: a specific purpose, a warning sign that tells people who is recording and why, and a retention period. The Board says footage for purposes such as detecting vandalism should in most cases be erased after a few days, and that keeping it beyond 72 hours needs stronger justification.
In the UK, the Information Commissioner's Office publishes detailed guidance on CCTV and video surveillance for organisations, and police and local authorities in England and Wales must also have regard to the Surveillance Camera Code of Practice. In India, the Digital Personal Data Protection Act sets duties of notice, purpose limitation and security for personal data, which includes camera footage.
Workplaces and homes. Employers can usually use cameras for security and safety, but must tell staff, keep monitoring proportionate and stay out of toilets and changing rooms. UK guidance on domestic CCTV says that once a home camera captures neighbouring property or a public pavement, data protection law applies.
The practical checklist is short. Write down the purpose. Put up clear video surveillance signs. Set and enforce retention. Limit who can view and export. Redact bystanders before sharing. Assess any face recognition with particular care.
Every camera is a computer on your network, and an unmanaged one is an open door. The basics prevent most incidents.
Change every default password. Keep camera and recorder firmware current. Put cameras on their own network segment, away from office systems. Never expose cameras or recorders to the internet through port forwarding. Encrypt video and logins in transit. Use central sign in with your directory, give each user only the permissions their role needs, and review the audit trail. Our video surveillance cybersecurity guide covers each step.
Video surveillance cost depends less on the price list than on the licensing model, so compare models before comparing quotes. Total cost has four parts: cameras and installation, servers and storage, software licences, and the people who run it.
Per camera perpetual licences. Many traditional VMS products charge a one time licence for each camera or channel, plus an annual maintenance fee for updates and support. Costs scale directly with camera count.
Per camera subscriptions. Cloud services usually charge a recurring fee for each camera, often tied to the vendor's own hardware. The fee never stops and grows with every camera.
Flat or tiered licences. Some platforms charge a flat fee per server or per plan, regardless of how many cameras connect. This rewards growth and makes budgets easier to defend.
AI pricing. Analytics may be bundled, sold per camera, or sold per stream. Ask how AI is priced separately from the VMS, because it can become the largest line over time.
Beyond licences, the biggest cost drivers are storage, installation labour and the hours operators lose to false alarms.
1. Define the risks and outcomes. List what you must prevent, detect and prove, site by site.
2. Survey the site by day and by night.
3. Decide what each camera must achieve: detect, recognise or identify. That sets resolution, lens and placement.
4. Inventory what you already own. Existing cameras and NVRs can often stay.
5. Choose the deployment model your policy and network allow.
6. Choose the software, using the ten criteria above, before the cameras.
7. Design the network and storage: PoE budgets, bandwidth, segmentation and retention days.
8. Complete privacy work: purpose, signage, retention and, where needed, a data protection impact assessment.
9. Install, then tune analytics and rules with real walk tests.
10. Train operators, then review alarms monthly and adjust. Our scalable video management system guide covers growing beyond the first site.
Visylix is an automated AI video surveillance platform built in India. It does not just record and display video. It watches, decides and acts.
It watches. 22 AI analytics are built in house: face recognition with watchlists and operator confirmation, crowd detection, falls, tailgating, number plate recognition with allow, block and watch lists, intrusion, line crossing, weapons, fire and smoke, PPE, abandoned objects, camera tampering, anomaly detection and audio events such as gunshots. Each camera learns its own normal, and marking a false alarm tightens thresholds automatically, never below the operator's limit.
It decides. Sudarshan, the Visylix automation layer, overlays live AI on the focused feed and builds rules with AND and OR conditions, ordered time sequences, branching actions and video plus access control correlation. Rules can be tested on real history before they go live.
It acts. Alerts reach the right team, alarm incidents open in the operator console, and signed webhooks, MQTT, SIEM and PSIM integrations carry events onward. Talkdown speaks through the camera speaker, and siren and door strike relays respond physically. Radha, the AI copilot, runs on an on premise language model with 111 tools: ask in English, Hindi or Hinglish to search footage, follow a vehicle across cameras or write an automation. High impact actions always ask for confirmation, and every call is audit logged.
It works with what you own. ONVIF discovery, direct RTSP, ONVIF device packs for the top five brands, and playback from existing Hikvision, Dahua, Uniview, I8H, Bosch and Panasonic recorders. Analogue cameras join through the DVR they are already wired to.
The control room is complete. Sub second live view over WebRTC in any modern browser, down to under 200 ms in ideal conditions. Video walls up to 6x6, floor plan and geographic maps, PTZ patrol tours and autotracking, Active Directory and LDAP sign in, 46 permissions across 4 roles, access control connectors and 55 interface languages.
Evidence holds up. Locked footage is sealed with SHA-256 and exported in a player that opens offline, so anyone can check it is untouched. Faces and plates are redacted on export.
Your video stays with you. Visylix ships as a Docker image on your own hardware: on premise, at the edge, hybrid, fully air gapped with an offline licence, or in your own cloud account. A standby server takes over through an operator cutover in under 60 seconds.
Plans. Starter at ₹4,999 or $49 per month and Pro at ₹9,999 or $99 per month include unlimited cameras on the VMS licence. AI analytics are a separate subscription billed per camera per month, on the cameras you choose. Talk to us for AI pricing. Enterprise is custom, and a free 7 day trial is available.
Video surveillance earns its place where there is something worth protecting and someone ready to respond: perimeters, entrances, car parks, warehouses, retail floors, campuses and public spaces; as evidence that holds up when an incident is disputed; as a deterrent that is visible and signed; and, with AI and automation, as a system that spots events and acts on them faster than any operator watching a wall.
It disappoints when cameras are bought before the risks are defined; when nobody watches the alarms or tunes the analytics; when footage is kept without purpose; when cameras sit unpatched on the open internet; when licence fees grow with every camera added; and anywhere a camera intrudes on privacy more than the risk justifies.
Video surveillance goes by several names. The oldest is CCTV, short for closed circuit television, from the days when analogue cameras fed a private set of monitors. Other common terms are video security, video monitoring and a video surveillance system (VSS). The software that runs a modern system is called a video management system, or VMS. All of these describe the same idea.
Video surveillance means using cameras and software to watch, record and analyse activity in a place, in order to protect people, property and operations. It covers live monitoring by operators, recording for later review and evidence, and, increasingly, AI analytics that detect events automatically and trigger a response. It applies to homes, businesses, campuses, cities and critical infrastructure alike.
In practice there is very little difference, and most people use the two terms interchangeably. Where a distinction is drawn, a security camera is often a single device protecting one home or shop, usually with its own app. A surveillance camera is one part of a managed system, with central recording, monitoring, analytics and an audit trail. The camera hardware can be identical.
Video surveillance software, usually called a video management system or VMS, is the application that connects cameras, records their video, shows live and recorded footage to operators, and manages users and permissions. Modern video surveillance software also runs AI analytics, applies alarm rules, connects to access control and other systems, and exports evidence. It is the part of the system that turns cameras into a working security operation.
It depends on where you work, and this is general information, not legal advice. In many countries an employer may use cameras for a clear, legitimate purpose such as security or safety, but must tell staff, keep monitoring proportionate, and never film private areas such as toilets or changing rooms. European guidance treats constant monitoring of employees as a serious intrusion. Check your local law and your employer's written policy.
Usually a camera that covers only its owner's own property is lawful, but rules vary by country and state, and this is not legal advice. In the UK, guidance on domestic CCTV says that once a home camera captures neighbouring homes, gardens or public pavements, data protection law applies, and the owner must minimise intrusion and respect privacy rights. Recording audio is often more tightly restricted than video.
The best video surveillance system for a business is the one that works with the cameras it already owns, scales to every site, shows live video without delay, and uses AI to detect and act on real events. It should keep evidence tamper evident, control who sees what, respect privacy law, and have a licence model that stays predictable as the camera count grows.
Only as long as the purpose needs. The European Data Protection Board says footage kept for purposes such as detecting vandalism should in most cases be erased, ideally automatically, after a few days, and that keeping it beyond 72 hours needs stronger justification. Many organisations set 30 days by default and lock specific clips as evidence for longer. Write the period down and apply it automatically.
Visylix is an automated AI video surveillance platform that watches, decides and acts. It brings 22 AI analytics built in house, Sudarshan automation rules, the Radha AI copilot, sub second live view in any browser and tamper evident evidence export. It works with the cameras and NVRs you already own and runs on your own hardware. Starter and Pro include unlimited cameras on the VMS licence.