From CCTV to operational intelligence: what the next generation of industrial video looks like, and why the cameras you already own are the place to start.
In brief
- Video is shifting from passive evidence to machine-readable operational data, used by Security, HSE, Operations, Maintenance, IT/OT and Management.
- People cannot be the detection engine for hundreds of feeds. Automated detection closes that gap only if it is engineered as an event workflow, not shown as a demo.
- Analog cameras still hold more than half of India's CCTV market, so the practical route is brownfield: assess, retain, add intelligence, and replace only where necessary.
- Start with the event you are currently failing to see early enough, not with an analytics catalogue.
A control room built for recording
For decades, industrial CCTV followed one architecture: capture, transmit, record, retrieve. A camera witnessed an event and a recorder stored it. An operator watched it live, or investigators pulled it afterwards.
That model quietly assumes a person is the detection engine, and people are poor at that job. A Sandia National Laboratories review of vigilance research found that the decline in detection performance is typically complete 20 to 35 minutes into a session, with at least half of the eventual loss occurring in the first 15 minutes. Load makes it worse: the decrement is more pronounced when observers must watch several displays for multiple critical events. Even trained staff struggle. In a controlled study of CCTV operators and novices in Applied Ergonomics, participants detected only about half of the target behaviours, and false alarms were high.
A video wall can show 256 feeds. No operator can watch them.
Why the stakes are rising
Globally, the ILO estimates that nearly three million workers die every year from work-related accidents and diseases, including about 330,000 deaths from work accidents, alongside 395 million non-fatal injuries. Asia and the Pacific accounts for 63 per cent of the global total, and the economic cost is estimated at around 5 per cent of GDP.
In India, the NCRB recorded 742 factory and machine accidents in 2024, resulting in 660 deaths. Accidents are becoming fewer but deadlier: about 66 deaths per 100 accidents in 2010, rising to nearly 89 by 2024. Oversight is thinning as exposure grows. Hazardous-sector employment has nearly doubled to 4.72 million workers, while inspection coverage of hazardous factories has fallen from 62% to 39%.
In refining, the scale is expanding. India has 23 refineries processing nearly 270 million tonnes a year, with capacity heading from 272 towards 300 MMTPA. In Marsh's 100 Largest Losses report, refining has the highest share of the 100 largest losses at 37, and 32% of all losses came from explosions and 24% from mechanical failures.
Video intelligence does not replace process safety, mechanical integrity programmes or gas detection. It adds a layer they do not cover: what people, vehicles, fire and smoke are doing across the site, continuously, at every camera.
Two recent Indian incidents show where that layer matters. Neither cause is established, and we draw no conclusions. In November 2024, a fire and explosion broke out in a vacuum distillation unit during start-up after a 40-day maintenance shutdown, and eight personnel suffered burn injuries. Start-ups and turnarounds bring changed configurations and many contractors on site. In June 2026, the operator said preliminary information suggested a naphtha pipeline fire occurred near an unauthorised theft point, and at least 20 people were reported injured. Long perimeters and pipeline corridors are hard to watch by eye at any hour.
The market is moving, and the numbers agree on direction
Estimates vary because analysts define "AI video analytics" differently. Every source, however, points to compound growth.
Six shifts reshaping industrial video
1. From CCTV to operational data
The traditional consumer of CCTV was the security department. A camera can now generate an event when PPE is missing, a restricted boundary is crossed, smoke appears, a vehicle enters a defined zone, crowd density changes or the camera itself is obstructed. Those events matter to HSE, Operations, Maintenance and Management as well.
Genetec's 2026 State of Physical Security survey (7,368 professionals across six regions) shows the direction. More than 70% of respondents use unified or integrated systems, and 60% say integrating new capabilities is their main reason for replacing legacy technology. For the first time, AI ranked alongside access control and video surveillance as a top project priority, and interest in AI more than doubled among end users. One caution: Genetec is a VMS vendor and its respondents were 42.5% channel partners and 37.5% end users, so treat the figures as directional.
Energy majors are already proving the model. ADNOC has deployed more than 500 AI-enabled RASED cameras for real-time hazard detection. Aramco's Salamatic Eagle combines CCTV feeds with machine learning to detect PPE misuse and safety hazards on drilling rigs and construction sites.
The trend is not "AI CCTV". It is video becoming operational data.
2. From rip-and-replace to brownfield intelligence
Industrial facilities have already invested in cameras, cabling, networks, storage and control rooms. In India, analog cameras still held 51.65% of the CCTV market in 2025, while AI-enabled cameras are the fastest-growing type at a 20.55% CAGR. A large legacy estate is not an obstacle to AI. It is the opportunity, provided you ask the right first question.
Not "which cameras should we replace?" but "which parts of the existing architecture can technically support the intelligence we need?" That means assessing resolution, frame rate, codec, protocols (ONVIF/RTSP), field of view, lighting, occlusion, network capacity and the operational use case. Brownfield AI should begin with an infrastructure assessment, not a replacement quotation.
3. From detection to event engineering
Demonstrating helmet detection is easy. Engineering an operationally useful PPE event is not. What confidence threshold counts as detection? Which zone applies, and does the rule run continuously or only during certain activities? How long must a condition persist before an event is created? What stops repeat alerts for one occurrence? Who receives it, does an operator verify it, when does it escalate, and what evidence is retained?
These questions separate a demonstration from an operating system. Poorly engineered analytics simply convert screen fatigue into alarm fatigue. The architecture that matters is:
Detect → Correlate → Prioritise → Verify → Notify → Escalate → Resolve
The objective is not maximum detection. It is actionable detection.
4. From a central VMS to edge, on-premise, cloud and hybrid
Cloud is not automatically the destination for every industrial video workload. A refinery, offshore asset or power facility may have legitimate requirements for latency, bandwidth, availability, cybersecurity, data residency and business continuity. Some analytics belong at the edge, video management or storage may stay on-premise, and other functions suit the cloud.
The market reflects this. In India-relevant analyst data, cloud held 57.89% of AI video analytics in 2025, but hybrid and edge architectures are forecast to grow fastest, at a 23.34% CAGR through 2031. Genetec's respondents likewise prefer hybrid cloud and want flexibility over where each workload runs. The architecture should follow the operational requirement, not the fashion of the moment.
5. From a security system to an enterprise platform
If video is only evidence, CCTV belongs to Security. If it generates information about worker safety, vehicle movement, access, emergency response and asset protection, the stakeholder map widens to HSE, Operations, Maintenance, IT, OT, Plant Management and Emergency Response.
Most organisations are not there yet. Genetec found that only 25% of end users actively share physical-security data with other departments. The value will come less from another surveillance dashboard than from connecting visual events to workflows that already exist.
6. From AI accuracy to AI governance
Connected, intelligent cameras enlarge the attack surface and the governance burden. Who can access the video? Where is it processed and where is metadata stored? Who can change an analytic rule? Are communications encrypted and operator actions logged? How long is footage retained, who owns the data, and can the organisation audit what happened?
These questions are moving into procurement. 70% of respondents in Genetec's research voiced concerns about how AI systems are designed and implemented, particularly around data use and understanding how the AI works. The next generation of industrial video cannot just be intelligent. It has to be governable.
The solution: SMEC NexView
SMEC NexView is an AI-powered video management and operational intelligence platform. It layers detection, alarm management and investigation tools over an enterprise video estate, in one environment instead of several disconnected products.
Analytics built for industrial environments
- Safety and compliance: PPE compliance, safety compliance monitoring, restricted-zone monitoring
- Security and perimeter: intrusion detection, virtual tripwire, loitering detection
- Emergency awareness: fire detection, smoke detection, crowd monitoring
- Operational analytics: vehicle monitoring, camera tampering detection, camera health monitoring
Event management, not isolated alerts. NexView correlates analytics events, device alarms and operational notifications, then prioritises them by urgency and routes them by email and mobile push, with a full audit trail on every event.
Recommended by LinkedIn
Command and control. Operators can view 1 to 256 cameras in flexible layouts, run video walls, and use PTZ presets and automated patrols from web or mobile. E-Map and GIS visualisation places cameras, alarms and zones on interactive facility maps, with one click to live video.
Investigation-ready evidence. Continuous, motion-, event- and schedule-based recording, smart search across cameras, synchronised multi-camera playback, and secure exports with chain-of-custody archiving.
Governable by design. Role-based permissions, multi-layer sign-in verification, encrypted communications, and a tamper-evident record of who accessed what and when.
Built for existing and future infrastructure. It works with cameras already on site through standard camera protocols, and integrates with access control, fire alarms, intercoms, building management and business software. It deploys on-premise, in the cloud or as a hybrid, scales without downtime, and fails over automatically so monitoring does not stop. Native Android and iOS apps give authorised teams live video, playback, alarms and full PTZ control from a phone.
Where the use cases emerge
The architecture is common. The operational problem is not.
The principle holds across all of them. Do not deploy an analytic because it exists. Deploy it because a defined operational event needs earlier detection or better verification.
Before you invest: audit these ten things
Only after these are answered does choosing analytics make sense.
In the field: NexView at a major Indian oil and gas operator
The starting point. The client, a major Indian oil and gas operator, already had CCTV covering its warehouse department. The cameras recorded, but the system was passive: spotting an event depended on someone watching a screen or searching recordings afterwards. The operator chose to convert its existing CCTV infrastructure into an AI-powered video analytics and VMS solution instead of replacing it.
What SMEC deployed. NexView now monitors 22 CCTV cameras connected through ONVIF and RTSP integration. The architecture is compact by design:
- 2 VMS servers in a High Availability configuration
- 1 analytics and storage server running AI event processing
- A centralised NexView dashboard, with SMS and Gmail alerting
Six analytics, chosen for a warehouse. Intruder detection, motion detection, object-missing detection, abandoned-object detection, camera-tampering detection and crowd counting. Object-missing detection adds a layer of protection for warehouse materials, equipment and assets. Tampering detection protects the integrity of the surveillance system itself.
An alarm philosophy, not just detection.
- Schedule. Each analytic has its own active time window, including holidays and weekends. Intrusion detection, for example, can stay off while staff are present and switch on automatically after working hours, so anyone detected afterwards is flagged as a potential intruder. This cuts false triggers during normal operations.
- Validate. Detections pass through an event-validation step before evidence is generated.
- Evidence. Every confirmed event becomes a single record: snapshot, 10-second video clip, camera information, event type and timestamp.
- Notify. Alerts go out through the dashboard, SMS for immediate notification, and email with the snapshot and clip attached, so an event never depends on a single channel.
- Investigate. Operators can view the snapshot, play the 10-second clip and review event history from the dashboard, without searching through long recordings.
Built so monitoring does not stop. If VMS Server 1 fails, the HA mechanism promotes the hot-backup VMS Server 2 to the active role and operations continue. A single VMS server failure no longer takes surveillance down.
Governed by design. Role-based access control, encryption, secure authentication, audit logs and activity logs are all in place.
What this deployment shows
- All 22 cameras now generate events. Coverage moved from watching to detecting, with less reliance on continuous manual monitoring.
- Investigation starts from evidence. Each alert arrives with a snapshot and a 10-second clip, so security staff can assess what happened before, during and after an event without scrubbing recordings.
- False alerts are managed by design. Scheduling limits intrusion detection to the hours when a person on site is genuinely unexpected.
- The existing camera estate was the foundation. NexView integrated through standard protocols, so the upgrade happened in the intelligence layer, not the camera layer.
Why SMEC
NexView is backed by more than 25 years of engineering experience in industrial automation, OT, electrical and instrumentation, and oil and gas, marine and industrial domain work. That is the difference between installing software and delivering a working system in a live plant. SMEC brings brownfield integration capability, testing and commissioning support, training and lifecycle support, and regional delivery from India, the UAE and Singapore, with facilities in Cochin, Mumbai, Dubai and Abu Dhabi.
Don't start with a demo. Start with your CCTV estate.
If hundreds of cameras are already installed, don't assume hundreds need replacing. SMEC begins by assessing your existing camera estate, scene conditions, network, recording architecture, priority zones and integration needs. The assessment answers three practical questions:
- What can be retained?
- Where can intelligence be added?
- Where is modernisation actually necessary?
For new facilities, there is a bigger opportunity: engineer CCTV, VMS and AI intelligence as one architecture from day one, instead of procuring three disconnected layers and integrating them later.
Already running a large CCTV estate? Send us your existing architecture and three to five operational events your facility cannot afford to miss. That is enough to start the engineering conversation and the fastest route to a proposal.
Request a NexView Site Assessment Call: +91 8606047714 Email: sales@smecong.com
SMEC NexView | Video Management • AI Analytics • Incident Intelligence • Operational Awareness
Built for environments where missing one event matters.
+91 86060 47714 | sales@smecong.com
SMEC | Global Support
Explore SMEC
SMEC – The Solution Hub is our cross-industry knowledge series covering Oil & Gas, Marine, and Industrial systems where engineering, automation, and operations intersect.
Our Ecosystem:
India :
- SMEC Automation
- SMEC Technologies (AI Intelligence Solutions)
UAE :
- SMEC MARINE SOLUTIONS LLC
- SMEC NAVSOL
- SMEC Oil & Gas Solutions LLC
Training & EduTech
- SMEClabs
- SMEC Technologies
Our Digital Hubs:
- Corporate: www.smec.in
- Energy & Industrial: www.smecoilandgas.com
- R&D & Technology: https://smecrtd.com/
- Asset Intelligence: www.ProSet360.com
Join the Conversation:
- YouTube
