Project title: Automatically analysing faults in sewer networks using CCTV footage
The majority of the wastewater networks in the UK are too small for manual inspection, instead CCTV surveys are needed.These surveys require a CCTV camera to travel the
length of the pipe, whether attached to a semirigid wire or remote controlled PIG (pipe inspection gadget), so the collected footage can be analysed by a qualified engineer. This
process is both costly and time consuming. A system capable of automatically detecting faults, using only the raw CCTV footage would dramatically reduce time and costs of such surveys.
The proposed methodology would take the collected footage, analysing it frame by frame in order to extract segments of footage containing a fault. From there the fault can be further
classified, by either another program or qualified engineer. Reducing the duration of the footage from an order of days to only minutes will reduce the time required time to complete a survey and in turn the associated costs.
See Josh’s publication: Automated Detection of Faults in Wastewater Pipes from CCTV Footage by Using Random Forests
Josh is based at the Centre for Water Systems at the University of Exeter and collaborates closely with the University’s department for Computer Science. Currently working individually on this project, Josh is applying modern machine learning techniques in order to develop a tool capable of assisting in the identification of faults in wastewater networks. Josh’s areas of
interest include: machine learning, image processing and optimisation, and their application to real world problems.
Previous to beginning his PhD Josh completed his BSc in Computer Science and Mathematics at the University of Exeter, and continued to expand his studies in the Water Informatics Postgraduate School.
Keywords: Wastewater Networks, CCTV, Fault Identification and Classification, Machine Learning, Automation.