The number of forest fires has increased significantly in recent years due to the deterioration in the health of forests as a result of accelerated climate change around the world. Longer periods of drought and pressure from a variety of pests and diseases, for example the bark beetle, are damaging large forest areas in northern Europe. In addition to financial losses, these impacts are resulting in reduced carbon accumulation and major ecological changes.
In 2022, a total of 2,397 fires were recorded in Germany. More than 3,058 hectares of forest were destroyed, predominantly in the federal states of Brandenburg and Saxony. Monitoring forest areas with aircraft or unmanned aerial vehicles (UAVS or 'drones') can help detect potential problems early. Even the observation towers equipped with the latest in camera and sensor technology, combined with the Artificial Intelligence (AI) to detect plumes of smoke as an indication of fires, can be an ideal solution to help in the fight against the widespread destruction of wildlife habitats.
Forest monitoring project
A research project launched by the Landkreis Goerlitz in Saxony and the German Federal Ministry for Digitalization and Transport (BMDV), in collaboration with the company Geotechnology, Geoinformatics and Services GmbH (GGS), focused on the monitoring of a total of 170km2 of forest area in northeastern Germany. To carry out this project, various approaches were used. Given the extensive area covered, topographic planes and UAVs were used for aerial monitoring, while on the ground observation towers with state-of-the-art sensor technology and ground sensors were used.
Camera setup for aerial survey
The camera setup for the aerial survey was based on GGS OIS technology with the addition of near-infrared (NIR) and thermal cameras. GGS chose a Phase One iXM-RS 150 with a 90mm lens for the nadir image. Two additional Phase One Achromatic iXM-100 cameras with NIR filter (700-850nm and 750-850nm bands) and two 70mm lenses were added. For the generation of a 3D surface model, four Phase One iXM-100 oblique cameras, each equipped with an 80mm lens, were also integrated into the camera capsule. In addition to the nadir cameras, two additional thermal cameras were used to capture the entire footprint at lower resolution to determine the influence of climate microchanges on forest health.
Data processing and model generation
The RGB, CIR and NDVI data collected were initially processed with Phase One's iX Capture software, before being processed with specialized photogrammetric software to generate true orthophotos. The red edge data were generated after producing the orthophotos of both NIR bands and were refined by raster conversion between the two NIR orthophotos. RGB data from the oblique system (nadir and oblique) were processed with Skyline's Photomesh software package to generate a 3D model.
Applications in forest management
The resulting data models can be used in various applications, including:
- Forest health assessment: RGB, NIR, CIR, NDVI and red edge orthophotos can be used by forestry experts to detect high fire risk areas, areas suffering from drought or pests/diseases, such as bark beetle, or to assess forest status. soil health. These areas are then labeled as high risk and therefore have a higher priority for fire watching. Red edge detection in the 700-750nm band allows for faster and easier detection of health problems than NIR/CIR imaging alone. Therefore, it can be considered as a quick indicator of stress within trees.
- Emergency maps: RGB orthophotos can be used as part of the emergency navigation system for firefighters in combination with other information
, such as the accessibility of roads, lakes to access water, information on critical infrastructure and inaccessible areas. This information, for example a mosaic of real orthophotos, is part of a cloud-based GIS application. It is frequently updated with additional data, for example data collected by UAVs on the dimensions and dynamics of forest fires, as well as the associated risks to nearby settlements and infrastructure. This navigation application can also be used to coordinate and direct emergency responders. - 3D simulation: 3D data from oblique cameras can be used to generate a perfect 3D model of the area being processed with Skyline's Photomesh software. Using Skyline's TerraExplorer software, analyzes can be performed on the surface model, such as visual line of sight from observation towers and UAV flights optimized to detect fire settlements. Using different AI modules, the 3D model also allows the simulation of forest fires to verify whether fires are visible from remote points such as observation towers or from UAV flights at different heights and using different flight strategies.
Next steps
Currently, there are three observation towers available within 170km2 of forest studied in this research project. However, only one of them is equipped with a camera system, and that camera system is based on older technology. As a next step, an innovative gimbal-mounted camera system capable of capturing RGB, NIR, AC and thermal data will be installed on two of the three towers. The gimbal will scan the area at 270 degrees, while the cameras will capture images with 60% lateral overlap. The resulting data will be transmitted via 5G to a central crisis management server and analyzed using AI algorithms to detect wildfires. If a forest fire is detected, an alarm will be raised and a UAV will be deployed to assess the situation. Combined with observations from the other towers, the location of the fire can be easily calculated and verified.
Smaller areas can be monitored with UAVs equipped with RGB, NIR and red-edge cameras. To allow longer flight time, these drones only carry compact, small format cameras. After activating the alarm, a single UAV will initially be deployed. In certain situations, a swarm of UAVs can be sent to observe a larger area to search for other fires. The data collected, in combination with visual inspection of wind speed and direction, is crucial in determining the speed and direction of fire spread. Transmitted over the 5G network, the combination of all the data collected informs the strategy and deployment of emergency services, while minimizing risks to firefighters, settlements and infrastructure.
READ MORE ARTICLES ABOUT: Environment with AI.
READ THE PREVIOUS POST: Will 2024 be the Year of AI Governance?.