with the increase of frequency and strength of natural hazards (flash floods, major fires, hurricanes, tornadoes, etc.), the ability to monitor the state of our natural capital and its interplay with the human activity becomes essential and requires joint efforts from different fields and stakeholders. Particularly, the field of data science – including artificial intelligence (AI), remote sensing and computer vision – is becoming a cornerstone for environment monitoring. This thesis subject aims at developing novel AI and visual recognition solutions to assess different aspects of environmental changes and assist experts and operators getting observable facts. The ultimate goal is to support remote sensing (RS) based natural hazard monitoring with new AI methodologies and algorithms – that thoroughly integrate different sources of visual data and contexts (expert apriori knowleges, ground photography, etc.) – for different visual recognition tasks. The latter include assessing risks and damages before/after natural hazards for land-use planning and rescues in wide extents using high-resolution satellite and unmanned aerial vehicle (UAV) imagery. We pay a particular attention to extensively consider machine learning and visual recognition methods (namely deep and convolutional networks as well as long-short-term-memory-LSTM networks, etc.) in order to make them suitable to handle these types of data and their temporal evolution.
This PhD research project has been submitted for a funding request to “Sorbonne Center for Artificial Intelligence (SCAI)”. The PhD candidate selected by the project leader will therefore participate in the project selection process (including a file and an interview) to obtain funding.