Blog/Insights

AI Video Monitoring for Mining Sites: Haul Roads, Restricted Zones and Fuel Theft

AI video monitoring for mining sites flags haul road, exclusion zone, fuel bay and perimeter events on existing CCTV so teams verify and respond faster.

Mining sites cctv

AI video monitoring for mining sites adds a software layer to the cameras a mine already owns. It watches haul roads, exclusion zones, fuel bays, stockpiles and perimeter fences continuously, then flags configured events, such as a light vehicle entering a controlled area or after-hours activity at a diesel tank, so control-room staff can verify and dispatch. It complements, not replaces, onboard collision avoidance systems.

Key takeaways

  • Powered haulage remains a leading cause of mine deaths. It accounted for 13 of 33 US mining fatalities in 2025, according to MSHA data compiled by Pit & Quarry.
  • Most mines already have CCTV on gates, fuel bays and workshops, but few people watch it live. An AI layer turns that footage into reviewable events.
  • Collision avoidance systems (CAS) act on the machine in the seconds before a collision; video monitoring watches zones, compliance and theft across the site.
  • Diesel, copper cable and perimeter intrusion are recurring losses at mines in South Africa, West Africa and Latin America, with sourced costs running into the billions of rand.
  • Worker monitoring needs governance from day one: a stated purpose, a privacy impact assessment, worker consultation and a retention schedule.
  • Start with a 60 to 90 day pilot on 10 to 20 cameras, three or four event types and named owners for every alert.

What is AI video monitoring for mining sites?

AI video monitoring is software that analyzes live camera streams and surfaces configured events for human review. In mining, those events fall into four groups: vehicle and pedestrian safety, restricted-zone compliance, asset protection and perimeter security.

Kote by Kotelab connects to existing IP cameras and RTSP-capable NVR/DVR streams. For the mechanics of detections, zones and alert routing, see our guide to real-time threat detection with AI video analytics.

Key terms:

  • Haul road: the engineered road network that carries loaded and empty haul trucks between the pit, the crusher, stockpiles and waste dumps.
  • Exclusion zone: a defined area that people or certain vehicles must not enter without authorization, such as a blast area, a dump edge or an active loading face.
  • Light vehicle/heavy vehicle (LV/HV) interaction: any situation where pickups, buses or pedestrians share space with haul trucks, loaders or dozers.
  • Collision avoidance system (CAS): onboard equipment that detects nearby vehicles or people and warns the operator or intervenes on the machine.

Why are haul roads and vehicle interaction still the biggest safety risk?

In the United States, powered haulage was the leading cause of mining deaths in 2025, with 13 of 33 fatalities, based on MSHA figures compiled by Pit & Quarry (January 2026). When MSHA issued its 2023 rule requiring written safety programs for surface mobile equipment, it said powered haulage and machinery had been the leading causes of serious and fatal accidents in recent years (December 2023).

The International Council on Mining and Metals (ICMM) reported 42 fatalities across its member companies in 2024, 9 of them linked to mobile equipment and transportation (July 2025). In South Africa, the Minerals Council reported 41 mine fatalities in 2025, down from 42 in 2024 (February 2026), with transportation named as a focus area.

What does a haul road incident actually look like?

In a 2025 report from Western Australia's regulator, during a night shift, a light vehicle entered a controlled area where haul trucks were moving a stockpile, without radio contact with the operator in control (WorkSafe WA, March 2025). A loaded truck struck it and pushed it roughly 40 meters down a ramp. Nobody was physically injured.

One contributing factor the regulator listed: compliance with the safe system of work for that area was not monitored or enforced. A camera on the entry point, with a "vehicle in restricted zone" rule active during the shift, lets a supervisor see the breach as it happens rather than in the investigation.

How is AI video monitoring different from collision avoidance systems (CAS/PDS)?

A proximity detection system (PDS) or CAS is mounted on the machine. It detects people or vehicles close to that machine and either alerts the operator or acts on the machine. In South Africa, regulation 8.10 under the Mine Health and Safety Act was amended in December 2022 to require employers to provide systems that automatically slow or stop diesel-powered trackless mobile machines when the operator does not act (Mining Weekly, October 2023).

AI video monitoring uses fixed cameras to watch places rather than machines: haul road intersections, ramp entries, dump edges, fuel bays and gates. It does not stop a truck. It tells people that a light vehicle is in the wrong zone, that someone is in an exclusion zone without required PPE, or that there is activity at a gate after hours.

  • Where it sits. Collision avoidance / proximity detection (CAS/PDS): On each machine and often on personnel tags; AI video monitoring on fixed site cameras: On existing CCTV across the site.
  • What it watches. Collision avoidance / proximity detection (CAS/PDS): The space around one machine; AI video monitoring on fixed site cameras: Zones, roads, gates, fuel bays, perimeters.
  • Primary job. Collision avoidance / proximity detection (CAS/PDS): Prevent an imminent collision; AI video monitoring on fixed site cameras: Surface rule breaches, unsafe patterns, theft and intrusion.
  • Response. Collision avoidance / proximity detection (CAS/PDS): Operator alert or automatic machine intervention; AI video monitoring on fixed site cameras: Event with clip and context sent to people for review.
  • Covers theft and intrusion. Collision avoidance / proximity detection (CAS/PDS): No; AI video monitoring on fixed site cameras: Yes, where cameras cover the area.

A mine needs both. CAS handles the last seconds before a collision. Video monitoring catches the compliance drift that puts people in those seconds, and covers security risks CAS was never designed for.

Why do mines lose so much to diesel, copper and equipment theft?

Mines hold diesel, copper cable and portable equipment across sites that are hard to fence and patrol.

Copper cable theft is the removal of live or dormant electrical and communications cable for its scrap metal value. At a South African platinum industry event in March 2023, Sibanye-Stillwater said copper theft cost its PGM operations more than R1 billion in lost production in 2022, with security costs above R1 billion a year on top (Daily Maverick, March 2023). Our sister post on substation intrusion and cable theft detection covers the same problem on power networks.

In Chile, Carabineros data presented to a congressional commission showed recorded cable theft cases rising from 172 in 2021 to 461 in 2025 (Reporte Minero, August 2025), with prosecutors describing violent robberies at mining sites as one of the main patterns.

Fuel theft at mines ranges from small-scale siphoning to organized operations. In one 2023 case in Mpumalanga, South Africa, police arrested four men whose pickups had tanks fitted in their load bins and had hidden between stockpiles at a mine (IOL, April 2023). Zone and schedule rules at fuel bays and stockpiles catch this better than generic motion alerts.

What can video flag at a fuel bay or laydown yard?

Useful events include after-hours activity at the fuel bay, a vehicle in a restricted zone around bulk tanks, loitering near cable drums, and equipment tampering. Each arrives with a clip, so the control room can tell a scheduled refueling run from an unscheduled one before dispatching. Linking those events with gate badge data is covered in our guide to access control and video correlation.

How should mines treat illegal mining as a perimeter problem?

Illegal mining is the extraction of minerals without a permit, often on or under a licensed operator's concession. South Africa's minister of mineral and petroleum resources said the country lost an estimated R60 billion to the illicit precious metals trade in 2024 (SAnews, January 2025).

In West Africa, Reuters reported in 2025 that record gold prices were drawing more unofficial activity onto concessions, with Nordgold, Galiano Gold, B2Gold and Barrick Gold all seeing recent incursions (Reuters via TimesLIVE, June 2025). The same report said unregulated mining supplies about 30% of the region's gold and supports three to five million people.

Concessions are large and guards cannot be everywhere. Video monitoring flags perimeter breaches and after-hours activity near pits or stockpiles, giving protection services an earlier, verified picture. Escalation decisions stay with trained people and, where needed, the authorities.

Why is existing mine site CCTV under-used?

Most mines have cameras on gates, weighbridges, fuel bays, workshops and stores, but use the footage mostly after something goes wrong:

  1. Too many screens, too few people. One operator cannot meaningfully watch dozens of feeds for a 12-hour shift.
  2. Alarm fatigue. Basic motion detection triggers on dust, rain, wildlife and every passing truck, so operators learn to ignore it.
  3. Slow investigations. Finding the 40 seconds that matter across several cameras and a night of recording can take hours.

An AI layer narrows attention to configured events and makes footage searchable. With natural-language video search, an investigator can ask questions like "Was any light vehicle on the north ramp between 22:00 and 02:00?" instead of scrubbing timelines. For the staffing side, see AI agents to support CCTV monitoring.

What can an AI layer on existing mine cameras flag?

These Kote detection types fit mining, each configured per zone and camera:

  • Vehicle in restricted zone: a light vehicle on a haul-truck-only ramp or an unscheduled vehicle at the fuel bay.
  • Restricted-zone entry: a person at a dump edge, conveyor transfer or crusher zone.
  • PPE and restricted-zone safety: people in designated areas without required PPE.
  • Perimeter breach: fence or gate breaches around the concession, plant or explosives magazine.
  • After-hours activity: movement at stores, workshops or stockpiles outside operating hours.
  • Loitering and equipment tampering: someone lingering near cable drums or interfering with fuel tanks, pumps or cabinets.
  • Multi-camera monitoring: several views of one gate, intersection or fuel bay in one place.

Each event reaches the right team with site, zone, camera, time and clip, and people confirm, dismiss or escalate, with multi-site visibility for group security.

What about remote sites with limited bandwidth and power?

Many pits and exploration sites run on constrained links and generator or solar power. Ask any vendor how much bandwidth each analyzed stream needs, what happens to detection and recording when the link drops, and how events are delivered once it returns. A camera estate review, part of Kotelab's onboarding approach, shows which well-powered, well-connected cameras to start with.

How should mines govern video monitoring of workers?

Monitoring haul roads and plant areas captures employees and contractors, so governance is part of the project. The UK Information Commissioner's Office (ICO) guidance on monitoring workers is a useful reference even outside the UK: employers should carry out a data protection impact assessment, tell workers about monitoring before it starts, use the least intrusive means for a stated purpose, and keep data only as long as needed (ICO). Local law differs (POPIA in South Africa, GDPR in Europe, state surveillance laws in Australia), so involve legal counsel and worker representatives early.

Practical rules:

  • Document the purpose of each detection and do not repurpose it quietly.
  • Use safety events for coaching and fixing systems, not automatic discipline.
  • Set and apply a retention schedule (Kote supports customer-configurable retention).
  • Restrict who can export clips, and log chain of custody for evidence.
  • Review vendor data handling, for example Kotelab's Legal & Trust Center.

How do you run a pilot of AI video monitoring at a mine?

  1. Define three or four events. For example: light vehicle on a haul-truck-only ramp, after-hours activity at the fuel bay, perimeter breach on a known intrusion section.
  2. Name an owner for each event on each shift (pit supervisor, control room, protection services) and agree the next action.
  3. Review the camera estate. Pick 10 to 20 cameras with clear views, stable power and RTSP access through your NVR or VMS.
  4. Draw zones and schedules that match your traffic management plan, exclusion zones, shifts and blasting windows.
  5. Complete governance before go-live: impact assessment, worker communication, retention and access permissions.
  6. Run in review mode for two to four weeks, with operators labeling each event as valid, false or needs tuning.
  7. Measure verified events per week, time to verification and repeat breaches by location, then scale only where the pilot showed value.

Evaluation checklist: what should mines ask vendors?

  • Camera compatibility. Question to ask: Does it work with our existing IP cameras and NVR/DVR over RTSP?; Why it matters at a mine: Avoids a rip-and-replace across a large, remote estate.
  • Zone and schedule control. Question to ask: Can we set different rules by zone, shift and blasting window?; Why it matters at a mine: Mine conditions change hourly.
  • Routing and search. Question to ask: Can events go to different teams, and can investigators search footage in plain language?; Why it matters at a mine: Safety and security have different owners; investigations need speed.
  • Remote conditions. Question to ask: How does it behave on low bandwidth and intermittent links?; Why it matters at a mine: Common at pits and exploration sites.
  • CAS fit. Question to ask: How does it complement our existing CAS/PDS?; Why it matters at a mine: Avoids gaps and duplicated alerts.
  • Data security. Question to ask: What encryption and certifications apply?; Why it matters at a mine: Kotelab states AES-256 encryption and ISO 27001, with SOC 2 Type 2 in progress.
  • Privacy and retention. Question to ask: Can we configure retention and access? Is our video used for anything else?; Why it matters at a mine: Required for worker monitoring governance.

Frequently asked questions

What is AI video monitoring for mining sites?

It is software that analyzes live feeds from a mine's existing cameras and flags configured events, such as vehicles in restricted zones, perimeter breaches or after-hours activity at fuel bays. Each event arrives with a clip and context. Supervisors or control-room staff verify it and decide what to do.

Does AI video monitoring replace a collision avoidance system?

No. A collision avoidance system sits on the machine and warns the operator or intervenes in the seconds before a collision. Video monitoring uses fixed cameras to watch zones and roads for rule breaches, theft and intrusion. Most mines benefit from both.

Can we use our existing mine site security cameras?

Usually, yes. Kote connects to existing IP cameras and RTSP-capable NVR/DVR streams, including Hikvision, Dahua, Axis, Bosch and Hanwha Vision cameras. A camera estate review shows which cameras are worth connecting first.

How does video monitoring help with diesel and copper cable theft at mines?

It flags after-hours activity, vehicles in restricted zones, loitering and equipment tampering around fuel bays, cable drums and laydown yards. Security teams get a verified clip early enough to dispatch. Investigators can also search footage to reconstruct what happened.

Is it legal to monitor mine workers with AI video analytics?

It depends on the jurisdiction, so involve legal counsel. Common requirements include a clear purpose, a privacy impact assessment, telling workers before monitoring starts, using the least intrusive approach and keeping footage only as long as needed. Guidance such as the UK ICO's worker monitoring guidance is a useful baseline.

Put your mine's cameras to work

Kote by Kotelab turns the cameras you already have into real-time intelligence for haul road safety, restricted zones, fuel and cable protection, and perimeter security. We define the events and owners that matter at your site, review your camera estate, then configure and evaluate with your operators. For mine stores and logistics yards, see warehouse loading dock and cold room door monitoring.

Talk to Kotelab about a mining pilot.

Put your cameras to work

Intelligence for the moments that matter.