Project proposal: AI camera-based early wildfire detection system for 250 km² in NZ

Summary

The project plans to deploy AI-powered visual fire detecting and monitoring system across a 250km² forest area in NZ. It will utilize dual-spectrum cameras(combining visible light and thermal imaging) equipped with long-range monitoring and AI-based fire detection capabilities. The system will be supplemented by a hybrid architecture featuring 4G/5G wireless transmission and a LoRa sensor network.

The goal is to detect smoke and fire events in their early stages, transmit fire incidents in real time to the remote terminals, minimize response time, and reduce environmental impact, while maintaining a cost-effective, scalable deployment model suitable for remote terrains. 

Technical basis

AI-powered cameras have practical and outstanding capabilities to deliver all-weather, efficient, and reliable early-stage fire detection, forming the technological backbone of the system.

● Featuring HD resolution, long-focal-length zoom lens and high-accuracy-positioning PT with close-loop technology, SOARVISION AI PTZ cameras can detect fire/smoke over 10km away, delivering exceptional efficiency in large-scale environments.

● Combining visible light and thermal sensors enables accurate detection both day and night, even in a completely dark and harsh environment.

● Deep learning models can distinguish between real fire/smoke and false positives (e.g., fog, sunlight, dust), significantly reducing false alarms. 

● Digital signals can be flexibly integrated into management platforms, GIS, and monitoring centre, making them an ideal choice for automated early warning and monitoring networks.

With the SO977-TH-675A52 as the Recommended Camera, which is equipped with 4MP, 52x optical zoom camera and high-precision, large-angle PT, it can monitor a circular area of over 60 square kilometres during the day. In the evening, the 640×512 pixel thermal imaging camera with a 75 mm lens can monitor within a radius of 5-8 kilometers.

For a big forest covering 250 square kilometres, reasonably deploying 5-6 SO977 PTZ AI cameras can basically cover the entire area. At the same time, deploying an appropriate number of LORA smoke sensors in blank areas will significantly improve the cost-effectiveness of the project.

System Architecture

● SOARVISION AI Cameras (SO977, 5-6 units suggested, depending on the installation locations) – Fire/smoke detection and recognition with a 52x optical zoom camera, plus a 640*512 pixels thermal camera with 75mm lens.

● 4G/5G Industrial Routers (5-6 units) – Enables wireless video and data transmission to the backend or cloud server.

● Watch Towers/Mounting Poles – Provides an elevated, unobstructed location for camera installation, antenna, as well as a solar power system and battery.

● Solar Power Kits for cameras and telecom units (5-6) – Autonomous energy system per camera tower. Device power consumption is estimated 1.6-2 kWh/per average day, and battery requires 7 days storage capacity. Suggest: 800W solar panel + 20/25 kWh battery + MPPT controller.

● LoRa Sensors – Monitor temperature, smoke, and gas, transmitting lightweight alerts.

● LoRa Gateways – Receive sensor signals and forward them to the backend or cloud.

● Solar Power Kits for LoRa units.

● Hardware: PC terminal/server/NVR – Provide system management and storage.

● Software: SVMS Pro/Lite integrated security management platform – For equipment and video management, visualization, alert management, and data logging. It is a highly compatible, distributed video management system provided by SOARVISION, supporting storage devices from different brands. Alternatively, we also provide API or SDK access to other software platforms for integration.

● Cloud platform service is optional

Technical highlights

● The front-end camera is equipped with AI functionality that can detect/recognize fireworks, and can be set to transmit video only when an emergency is detected, thereby minimizing network load.

● Double checking capabilities with visible light and thermal imaging in all weather conditions.

● Cross-checking between the AI functions of the front-end camera and the back-end platform ensures more accurate alerts.

● LoRa Cost-effective total solution as Lora’s Long-range, low-power communication (~5–10 km per gateway)

● Platform-ready for future scaling (drones, satellite feed, etc.)

Expected Outcomes

● Real time early bushfire detection and emergency reporting, including fire location.

● 24/7 huge forest coverage with minimum human intervention