Summary
The “New Space” sector remodels the space industry. The multiplication of disruptive innovation methods has increased the possibilities offered in various fields: observation, telecommunication, bugging, navigation assistance, etc. Thus, video acquisition from space offers exciting perspectives of uses. The number of use cases could be numerous if the promises in terms of resolution, acquisition time or revisit are kept.
On one hand, the market of still video from space is embryonic but presents a potential of development as soon as operational systems offering satisfactory resolution and revisit capacities are available. On the other hand, Deep Learning algorithms provide performances never reached before by conventional algorithms. The possibility to improve spatial video data by these means allows to address new uses or new markets.
Indeed, these last few years, the Super Resolution topic on EO images has been tackled by Deep Learning techniques. The industrialization of these academic studies would allow to enhance even more the value of low-resolution and low-cost EO data, and thus to address new applications.
These emerging EO approaches have a tremendous potential, reaching a wide range of stakeholders, be they State, public or private. In order to validate the results of the study on real use cases, we have integrated in our consortium this wide panel of end users:
- DGA (French MOD) is interested in measuring the improvement in Detection, Recognition and Identification (DRI) metrics brought by SR,
- CEREMA, a public research centre focused on environment, mobility and land use, is interested to know if SR algorithms on space videos can facilitate illegal maritime activities detection,
- CEREMA is also interested in measuring port flows activities,
- HAROPA PORT, the leading port complex, is interested in how SR techniques on space videos can facilitate ship tracking in port areas to facilitate crisis management, implying ships involved in a collision, grounding or stranding on their way to the port,
- And ERAMET, a French multinational mining and metallurgy company having its own railway network, would like to know if SR algorithms on space videos can provide assessment on damages after a railway incident.
This study focuses on the implementation of these disruptive algorithms on the data provided by these new EO services. In order to assess the TRL of these techniques, the following work should be conducted:
- To carry out a state of the art review on the Super-Resolution methods addressed by Deep Learning techniques,
- To select and benchmark relevant SR algorithms,
- To prototype some solutions for the considered satellite dataset,
- To assess the results on concrete uses cases brought by real end users,
- To identify issues and to make recommendations for an operational implementation.
