TreeSatAI

Cognitive systems do not simply work by plug-and-play. Reliable input data is a fundamental requirement for Artificial Intelligence (AI). Up to now, the data preparation for AI approaches has required extensive human effort. We are therefore exploring numerous alternative sources for training data extraction, which could serve as highly reliable input data for AI.

The mining of highly valid samples, rather than collecting high numbers of less valid training data, is the central bottleneck. This paradigm shift poses a key challenge in satellite-based earth observation. Increasing volumes of free remote sensing data from different space missions as well as environmental geodata are becoming available for users. The boom of social media platforms such as Flickr and Open Street Map open up new possibilities to attain textual and visual information on the environment. Currently, the acquisition of high-quality training data is the biggest challenge for AI applications in earth observation, often leading to lower applicability to larger spatial domains.

The overall aim is the development of AI methods for the monitoring of forests and woody features on a local, regional and global scale. Based on freely available geodata from different sources (e.g., remote sensing, administration maps, and social media), prototypes will be developed for the deep learning-based extraction and classification of tree- and tree stand features. These prototypes deal with real cases from the monitoring of managed forests, nature conservation and infrastructures. The development of the resulting services by three enterprises will be supported by three research institutes.