Enhancing the Representation of WastewaterNetwork Maps Using Graphs

Abstract—Accurate representation of wastewater networks is critical for effective urban infrastructure management. Extracting these networks from low-quality geographical maps presents significant challenges due to incomplete or ambiguous information. So far, we have developed a method for extracting wastewater network structures from geographical maps and representing them as graphs. This method includes detecting key network elements, such as manholes, their identifiers (using Optical Character Recognition, OCR), and pipelines connecting them. As part of this approach, we developed an efficient algorithm to accurately associate manhole identifiers with their corresponding nodes, achieving acceptable results despite the low quality of image maps. To address the issue of isolated nodes caused by undetected components, we introduced weighted edges in the graph to quantify the likelihood of connections between nodes. This enhancement improved the representation of incomplete graphs. Our current research focuses on two key challenges: creating more complete and reliable graph representations of wastewater networks and detecting arrows that represent the direction of wastewater flow.
Index Terms—Wastewater networks, Graphs, Object detection, Geographical Maps.

*Ikram El Miqdadi and Fatima Abouzid contributed equally to this work.

ACKNOWLEDGMENT
This research has received support from the European Union’s Horizon research and innovation program under the MSCA (Marie Sklodowska-Curie Actions)-SE (Staff Exchanges) grant agreement 101086252; Call: HORIZON- MSCA-2021-SE-01, Project title: STARWARS (STormwAteR and WastewAteR networkS heterogeneous data AI-driven management). We would like to express our gratitude to ”Montpellier Méditerranée Métropole” and ”La  régie des eaux de Montpellier Méditerranée Métropole” for having provided us with data essential to this research.