@article{fdi:010095711,
  title = {{B}uilding a wastewater network graph and detecting anomalies from inspection videos},
  author = {{T}ran-{N}guyen, {M}.{T}. and {B}enferhat, {S}. and {C}hahinian, {N}an{\'e}e and {D}elenne, {C}.},
  editor = {},
  language = {{ENG}},
  abstract = {{T}he management of wastewater networks is essential to ensure optimal hydraulic modeling. {H}owever, data on wastewater systems, usually represented in {G}eographic {I}nformation {S}ystems ({GIS}), are often incomplete. {O}ther sources of information are thus required to complete them. {A}mong the potential sources, pipe inspection videos-traditionally used for maintenance and detecting issues, such as structural degradation, offer valuable, underexploited information. {T}his article goes in this direction and proposes to take advantage of these inspection videos to extract the structure of wastewater networks, represented here in the form of graphs. {F}or this, we propose automatic detection mechanisms of the different annotations associated with the inspection videos. {T}hese annotations contain enough information to construct a graph of the wastewater networks. {R}egular expressions and text recognition tools, based on optical character recognition ({OCR}), are used to extract manholes identifiers, water flow direction in pipelines, as well as other useful information to determine the geographical positions of wastewater network objects. {I}n addition to the topological reconstruction, effective management also requires assessing the condition of each pipeline segment within the graph structure. {T}o this end, we employ deep learning models-specifically, the {S}win {T}ransformer and {M}obile{N}et{V}2, for detecting anomalies and assessing pipe integrity. {E}xperiments conducted on real-world data validate the effectiveness of our approach, underscoring its potential to improve both wastewater network management and hydraulic modeling accuracy.},
  keywords = {{FRANCE} ; {PRADE} {LE} {LEZ}},
  booktitle = {},
  journal = {{SN} {C}omputer {S}cience},
  volume = {6},
  numero = {7},
  pages = {837 [17 ]},
  ISSN = {2661-8907},
  year = {2025},
  DOI = {10.1007/s42979-025-04372-9},
  URL = {https://www.documentation.ird.fr/hor/fdi:010095711},
}