Automated Detection of Defects/Anomalies and/or Critical Features Using Visual Data Acquired from Sewer Pipelines

Lucio Soibelman, Civil and Environmental Engineering

An ongoing research project is focused on automated detection of defects/anomalies and/or critical features using visual data acquired from a sewer pipeline infrastructure inspection and condition assessment process. Sewer pipelines are difficult to access by human inspectors; and usually they are difficult to be inspected and maintained using a human-entry approach .To reduce the risk of human entry inspection, video inspection (e.g. CCTV) is currently utilized. However, this inspection method is labor-intensive, time-consuming and error-prone. For instance, the miles of sewer being inspected yield hours of video that need to be processed by certified human inspectors. In this ongoing research, an automated and reliable defect detection capability is under development. To aid in automating this process, an automatic warning can be developed so that certified inspectors would not have to watch an entire inspection video during an inspection. The certified inspectors could be warned whenever a defect or a critical pipe feature has been found by the automatic warning system; and then they could pay attention to what is being automatically displayed. For the first stage of this research, this proposed automated defect detection methodology will be implemented for offline image/video-based condition assessment. The acquired visual data could be scanned and processed using the proposed detection methodology, and then defects/critical features would be framed automatically for certified professionals to further review and diagnose.

Work needed:

During the research validation stage, different visual data sets need to be tested and evaluated. The work needed is to develop a computer program capable of: preprocessing videos and images (e.g. capture image frames from videos, conducting image normalization and registration), feeding images into the developed automated detection algorithms (written in Matlab), and monitoring and testing the detection performance. You will work with professors Lucio Soibelman and Jim Garrett, both of whom are faculty in the Department of Civil and Environmental Engineering and members of the Center for Sensed Critical Infrastructure Research (CenSCIR), and their graduate students to complete this research.