Decision-Support System for the Rehabilitation of Deteriorating Sewers

Abstract

This paper describes an automated and integrated detection, structural assessment, and rehabilitation method selection system for sewers based on the processing of video footage obtained by closed circuit television surveys. The system is based on a neural network classifier (NNC) trained to identify longitudinal cracks in sewers. Results obtained from experimentation with the NNC indicate that crack detection based on single-frame processing is not sufficient, and frame sequence processing substantially improves crack recognition rates. Based on the location of the cracks, local and global structural damage is assessed and a rehabilitation method is selected. Based on the significance of damaged sewers, the rehabilitation projects are being prioritized. An expert system coordinates the various modules in the system and connects them to a geographic information system.