Identification of dead trees using artificial intelligence based on Earth remote sensing data

On the territory of the Russian Federation, there are massive shrinking of the stand, which are caused by the effects of pathogens, insects, fungal infestations, as well as the spread of bacterial diseases of trees. The development of remote monitoring systems for forest ecosystems is extremely important for the management of forest resources and making informed decisions regarding the conservation and restoration of forests. The purpose of this study was the development of a software module for automating the identification process of dead trees based on Earth remote sensing data. The analysis of various detection and segmentation strategies, including traditional computer vision methods and neural networks based on artificial intelligence, made it possible to choose object detection as the main method due to its effectiveness in labeling and the ability to quantify areas of tree damage. The combination of the object detection method with high-resolution images obtained using UAVs proved to be the most effective for accurate detection of dead trees.