Comparative Analysis of Wildfire Segmentation Using Copernicus Sentinel and Nimbo Earth Platforms Multispectral Images

Wildfires have a significant impact on the environment and necessitate effective monitoring and management strategies. This study uses satellite imagery to analyze and compare wildfire segmentation techniques using data from Copernicus Sentinel and Nimbo Earth platforms. Satellite images of five distinct wildfires were downloaded, including near infrared (IR), RGB, and Normalized Difference Vegetation Index (NDVI) images from both sources, resulting in a total of 30 images. A manual segmentation of one representative image per wildfire event was performed to serve as the ground truth. A customized mean shift algorithm was applied in Python to segment each of the downloaded images. The segmentation results were then evaluated against the ground truth images using the Dice Coefficient as a metric. The Copernicus Sentinel-2 provided high-resolution images with spatial resolutions of 10 meters for visible and near-infrared bands. In contrast, the Nimbo Earth platform offered mediumresolution imagery at 10 meters per pixel but with the advantage of providing up-to-date, cloud-free images on a monthly basis. The results indicated that the highest Dice Coefficient was achieved on the infrared images from the Copernicus Sentinel, demonstrating superior segmentation accuracy. This highlights the importance of using high-resolution infrared images for precise segmentation and monitoring of wildfire spread. In addition to using currently available IR, NDVI, and RGB images, future researches could explore the potential of other multispectral images, the integration of other advanced segmentation algorithms and multiple data sources to enhance wildfire analysis.