< Back to previous page

Project

Deep Sewer Mapping: A deep learning approach to generate digital twins of sewer infrastructure using low-cost sensors

Maintenance of aging sewer infrastructure is a critical challenge for municipal administrations. The goal of this project is to combine low-cost 3D cameras and cutting-edge AI technology to create a Digital Twin model of sewer systems enabling both 3D mapping of structural elements (i.e. pipes network) and automatic defects detection (i.e. cracks, deformations, infiltration, deposits). Nowadays, sewer inspection is already carried out with the help of remote camera systems, which allows to inspect sewer infrastructure in an organized and safe way. These cameras are capable of inspecting even the tightest sewer pipe and broadcast the images to ground operators for visual interpretation. However, there are limitations to current approaches which this project will allow to overcome: (1) visual interpretation of the imagery is a time consuming task and prone to human error; (2) the imagery and its interpretations are not properly georeferenced; (3) there is no full 3D reconstruction and mapping of the sewer infrastructure. This project will develop an integrated solution to acquire images and point clouds data of sewer pipes and a fully automated software to process the raw data and generate a geocoded Digital Twin model of the infrastructure, useful for inventorying and visualizing the current state, and planning further maintenance/construction operations. 

Date:1 Oct 2021 →  30 Sep 2023
Keywords:sewer inspection, digital twin, underground mapping, low-cost sensors
Disciplines:Surveying, Infrastructure engineering and asset management, Machine learning and decision making, Geomatic engineering not elsewhere classified