Filming the Earth in relief: 3D reconstruction and change detection – Filmer la Terre en relief
The goal of the project is to define the complete algorithmic chain that, starting with sets of stereo pairs or triplets of images of the Earth obtained by satellite, recreates the evolution of the ground relief at high resolution, detects its significant changes, gives by fusion the most plausible relief a each time, and finally projects back on this common 3D reference all images taken by satellite, merging them when they are synchronic, and equalizing them for a better time comparison when they are diachronic. The result will be what we shall call a “movie of the Earth”. Its use is contemplated at two very different scales. In the immediate future we will of course not be aiming at a global coverage. The tool should be able to respond to requests localized in time and space. The goal is that public and private operators get an automatic answer to any query : “see what happened”, on a fixed time range and on a specific site.
The advantage of satellite imaging in this field is threefold : it allows access at any time and in any area without geographical or political restrictions. The rapidly increasing number of satellites promises daily coverage of any landmark. Finally, the observation of the Earth has a history of forty years, and the “Earth” movie will permit a resumption of images acquired in this period. An objection to the project is the excessive price of a “movie of the Earth”, considering the significant price of each satellite image. But this price is expected to decline very quickly thanks to the massive use of these data when they democratize, through channels including internet. The reason why we’re launching this project with a great economy of means, is that we think we can do it. Indeed the CMLA team has been working for fifteen years with CNES on the reconstruction of the relief of the Earth and on the design of stereoscopic satellites. This collaboration has led the team to produce two years ago a first fully automatic chain processing stereo pairs and triplets, S2P ( it satellite stereo pipeline).
We detail in this project the tools that will permit to expand this chain into a diachronic monitoring of the Earth, and we give evidence of its feasibility. The important thing is that the whole chain should be completely automatic, and therefore allow a diachronic synthesis for relief Earth on demand and within a reasonable time, in particular for military and civil uses requiring the detection of significant relief changes and a common visualization for all images of the same site.
This is particularly important for the damage assessment after disasters (landslides, floods, earthquakes), in the stages of military operations, in the analysis of the evolution of sensitive sites, and finally for all economic and organizational applications where the monitoring of the evolution of the construction of large areas is required. The team gathered around CMLA can do that in three years using the current dynamics of the team and its existing partnerships. Three key partners are involved in the project, by funding other parts of the project and by feeding data to it : CNES, SERTIT, and CESBIO. The requested contribution will fund the work of a postdoc and a research engineer for three years.
Monsieur Jean-Michel Morel (Centre de Mathématiques et Leurs Applications)
The author of this summary is the project coordinator, who is responsible for the content of this summary. The ANR declines any responsibility as for its contents.
CMLA Centre de Mathématiques et Leurs Applications
Help of the ANR 299,592 euros
Beginning and duration of the scientific project: December 2017 - 36 Months