Acquiring and processing the data needed to maintain underground water, gas and electricity networks is a costly and time-consuming process. Geneva's Industrial Services decided to call on HEIG-VD's Institut d'ingénierie du territoire to help them explore the impact that AI and new technologies could have on the acquisition and processing of cartographic data.
Services Industriels de Genève (SIG), responsible for the distribution of water, gas and electricity in the canton of Geneva, faces a major challenge: keeping its underground network cadastre up to date. As part of a process of data governance and technology watch, GIS decided to explore solutions integrating AI and new technologies, which would enable them to improve the acquisition and processing of field data. They called on HEIG-VD's Institut d'ingénierie du territoire (INSIT) to help them in this endeavor. The team was quickly assembled with Fabio Mariani from GIS, Professor Adrien Gressin from INSIT, and expanded to include Professor Giovanna Di Marzo Serugendo from the University of Geneva. Professor Gressin also called on Claude Müller, Alliance's innovation advisor, to support them in putting together the project's financing package.
Innovation through photogrammetry and artificial intelligence
To maintain distribution networks, GIS digitizes these underground networks on construction sites. This task, traditionally time-consuming and costly, requires the intervention of surveyors in the field. Surveys are carried out on site, and can become extremely complex due to the density of the various intertwined networks. The data acquired is then used to create the maps needed to maintain and develop the networks. The question at the outset of the project was whether it would be possible to automate the processes, so that a person already present on site could collect and receive the data analysis, and how by extension the surveyor's profession could evolve.
The project was carried out in several phases. The GIS network cadastre geomaticians began by collecting the data required for the project, capturing thousands of photos. This was followed by a second phase, in which Professor Gressin of INSIT processed the 2D images acquired in the field and reconstructed them in 3D. Then, still at INSIT, a 3D information extraction algorithm was used to reconstitute a 3D model. In the next phase, the 3D model was semantically checked by Professor Di Marzo Segundo of UNIGE to verify the consistency of the objects. Finally, the 3D model was verified in the field.
The project's two main innovations lie, on the one hand, in the joint use of AI and photogrammetry to create a labelled 3D model from several images of the same object, from different angles; a process that goes beyond the simple detection of an object on an image. On the other hand, the innovation also lies in the semantic reasoning implemented by UNIGE, i.e. the verification that the objects thus obtained respect the construction rules of the network.
Alliance coaching to obtain Innosuisse financing
Claude Müller, Alliance's innovation consultant, was with the project from the outset. Called in by Professor Gressin, he provided support from the drafting of an initial summary to the preparation of the Innosuisse dossier. During the various coaching phases, his expertise enabled him to ask the right questions, help set up the coordination and organization of the players involved, and refine the technical and financial arguments to be put forward in the application. The latter has now been submitted to Innosuisse and will be accepted.
If Claude Müller hadn't helped us formulate our application, we might not have received Innosuisse funding.
A fruitful collaboration set to continue
The project went well, with very good results in the case of pipe networks. The transfer of knowledge between INSIT and GIS has been completed, and the latter are now entering the industrialization phase. However, while the results are convincing for pipes, cables - through which street lighting, for example, passes - are cast in layers of material, such as concrete or sand, hiding certain parts of the network and preventing image analysis. "This will probably lead to another project focusing on this more specific type of network," explains Fabio Mariani. New collaborations are therefore on the horizon between GIS and INSIT.
What's more, tools are evolving fast, and as part of a technology watch process, we need to keep abreast of the changes these innovations bring. This interest in the evolution of the profession is shared within GIS, not only by the project leaders, but also by the geomaticians who took part in the data collection and enthusiastically apprehended the possible transformations and the future of their profession.
INSIT, for its part, has already automated the process, and this project opens up new applications and prospects. During the course of the project, INSIT developed its skills in machine learning, deployment and industrialization. And they can now envisage this type of processing on the scale of a territory, for example, to be able to better qualify the data. A fruitful collaboration, which has also enabled the institute to refine its expertise.
As a Haute École, our role is also to work with companies, which is very enriching, because these are real social issues.
A success story at every level
If the project is a success from the point of view of technological innovation, it is also a success from the human point of view. The understanding between the partners has been excellent, and the three teams from SIG, INSIT and UNIGE have progressed together, developing a collective intelligence over the months. "We were able to rely on GIS for project management, and they trusted the schools to bring in their skills and implement the methods, which was very appreciable," explains Adrien Gressin. "It's a real success story, both technological and human, and that's rare," enthuses Fabio Mariani.
The Institut d'ingénierie du territoire (INSIT) in brief
Located at: Haute École d'Ingénierie et de Gestion du canton de Vaud (HEIG-VD), Yverdon-les-Bains
Skills: Geomatics, planning and development, construction
Website: heig-vd.ch/rad/instituts/insit
Services Industriels de Genève (SIG) in brief
Located in: Geneva
Skills: Electricity, gas, drinking water and wastewater, thermal engineering, waste-to-energy and fiber optics
Website: sig-ge.ch