21:24, 01 March 2026

In Khakassia, AI Video Systems Spot Wildfires From the Sky

Russian regions are rolling out AI-powered video monitoring and deploying drones to prepare for wildfire season.

Authorities in the Republic of Khakassia are gearing up for wildfire season by purchasing dozens of drones to monitor natural fires. At the same time, officials are expanding a neural-network-based video analytics system that processes footage from drone-mounted cameras and ground sensors, helping dispatchers detect suspicious events more quickly. Anton Frolov, development director at RVi Group, outlined how the system works.

“This is a network of cameras with neural networks, various sensors, and intelligent software capable of drawing conclusions even from limited data. Unlike a human observer or conventional sensors, it can detect signs of ignition at the earliest stages. That allows a dispatcher to alert emergency services more quickly, gaining critical time to contain a fire. This is especially relevant for Khakassia, where 1,610 fires occurred in 2024, resulting in injuries,” Frolov said.

Cameras, Drones, and Satellites Tackle Blind Spots

Because the region includes vast sparsely populated areas, installing cameras everywhere is not feasible. Instead, Khakassia has built a hybrid monitoring model in which digital tools augment human oversight. The region has deployed a network of 60 pan-tilt-zoom cameras with 30x optical zoom. A single camera installed at a height of 70 to 100 meters in a fire-prone zone can provide coverage within a radius of up to 30 to 40 kilometers, helping monitor the most vulnerable forest areas.

The video analytics platform functions as a filter for operators. Humans cannot continuously watch dozens of screens, so the smart cameras automatically flag suspicious incidents and prioritize the most critical frames for dispatcher review.

Quadcopters are used to inspect lowlands and gorges. The Khakassia Forestry Ministry currently operates 34 drones and plans to expand the fleet. Russian-made drones are integrated into the regional early warning system: a neural network analyzes video streams in real time, automatically identifies thermal anomalies, cross-references them with camera footage and sensor data, and sends an alert to dispatchers when a potential threat is detected.

During the 2025 wildfire season, only 140 hectares of forest land in Khakassia were affected by fires, well below the allowable threshold of 3,500 hectares.

Territory management and ecology