We are currently experiencing a digital revolution where the traditional profession of cartographer has become digitalized. As a result, 3D geospatial maps are now indispensable in many emerging sectors such as connected cities (smart city), autonomous vehicles and economic intelligence...
What is a digital twin?
A digital twin is simply a virtual representation of a more or less complex object. The main interest is to be able to study the latter and especially to combine artificial intelligence and deep learning in order to create digital simulations.
In the era of Big Data, with the availability of huge volumes of remote sensing data and new technological tools, the automatic production of large-scale 3D geospatial data is now possible. In order to guarantee the optimal quality, this digital transition requires some manual interventions.
It is with these new technologies adapted to3D imaging that the AI4GEO program aims to develop a complete solution to automatically produce large-scale 3D products with the final objective of a global digital twin.
Digital twins and the city
The fusion of several data sources is the challenge of the cartographer:
- To succeed in gathering the data we have to reach a precision never reached before with the precious help of algorithms
- Deep learning
By combining these additional sources of information, there is an opportunity in the future to extend the participation of these digital twins to a wide range of services such as smart cities. A level of detail that is being closely studied to enable mapping of buildings and infrastructure into multiple levels of detail (LOD).
Research in the service of artificial intelligence
AI4GEO is a research and development program aiming at automating the production of real-time 2D and 3D geographic data. It is led by a collaborative and productive consortium of academic (CNES, IGN, ONERA) and industrial (CS Group, AIRBUS Defense and Space, CLS, GEOSAT, QUANTCUBE) partners, all experts in Geospatial Information.
The research project consists in developing the foundations for the automated production of very high resolution digital 3D maps and their additional information layers. Thanks to this solid study, a new data analysis platform designed to access, manipulate, process and visualize at scale all types of geospatial products is implemented.
Moreover, with the help of advances inimaging, AI and Big Data technologies as well as an unprecedenteddatabase, this collaborative project will be well equipped to achieve its goals.
Smart Cities, autonomous transport and economic intelligence
AI4GEO is doing its utmost to guarantee the excellence and the digital continuity of its action, as well as to collaborate on future topics such as smart cities, environment, autonomous cars, digital platforms or economic intelligence. However, this list could be extended in the future to other topics such as :
- Machine learning
- The Internet of Things (Iot)
- Virtual reality, etc.
Automatic 3D maps
The four-year project aims to automatically produce very high resolution 3D maps on a global scale. The digital transformation is underway, a complex model has been developed and deployed in the platform during the first two years.
This first version aims, by coupling the digital surface model and satellite images, the extraction of data on buildings and the first 3D reconstructions. Our algorithms are mainly based on supervised deep learning. Two softwares are developed (Labelcooker and Autolabel) to increase the productivity of the labelling of digitized objects, always with the aim of improving the accuracy of the digital twin to the maximum
To conclude, during its second year, the AI4GEO consortium has progressed
- Creation of a global digital twin
- Performance optimization
- Expansion to a wide range of cities and regions around the world
- Multi-resolution classification
- 3D mesh reconstruction
- Automation of the complex processing model
- High scalability to meet the global scale objective