The University of Sheffield
Browse
F16.pdf (418.39 kB)

CCWI2017: F16 'Automatic Detection of Sewer Faults Using Continuous CCTV Footage'

Download (418.39 kB)
journal contribution
posted on 2017-09-01, 15:02 authored by Joshua Myrans, Zoran Kapelan, Richard Everson
The work presented in the paper demonstrates the automatic detection of faults in sewers using CCTV footage. Specifically, it develops the fault detection methodology for application to a continuous video sequence, implementing a new ‘Smoothing’ stage. This new stage implements a Hidden Markov Model and Windowing technique, able to incorporate information from an entire CCTV segment to improve the frame by frame identification. The modified methodology was tested and demonstrated on footage taken from real world surveys in the UK. The results obtained, demonstrate that the extended methodology has improved performance with up to an 8% increase in accuracy and up to 5% increase in area under the ROC curve. Given the ability to produce predictions comparable to those of surveyors, the methodology shows promise for future application in industry.

History

Usage metrics

    Department of Civil and Structural Engineering

    Licence

    Exports

    RefWorks
    BibTeX
    Ref. manager
    Endnote
    DataCite
    NLM
    DC