Abstract
A framework for real-time and automated monitoring of the condition of culverts based on identifying internal defects via video inspection is developed. Manual inspection using closed-circuit television has several drawbacks, such as inconsistency and subjectivity due to human evaluation. Analysis of digital video, consisting of thousands of megabytes even at lower resolutions, can be laborious and not suited for real-time implementation. An innovative approach is to extract image frames judiciously from the video and analyze frames to locate and categorize major defects. Rather than analyze all extracted frames, one can skip consecutive frames at a minimal loss of accuracy and bring considerable savings in memory and system requirements. Each frame is preprocessed to enhance contrast, through use of an adaptive scheme, and to reduce dimensionality in pixel-space by implemention of region-based processing. Preprocessing is followed by a two-step image segmentation process, which implements a background elimination procedure in the first step and shape detection in the second step. Fuzzy clustering is used as the underlying segmentation model. Defect shape and depth information after post-processing are used as input to an assessment methodology for automated condition state. A simple formulation based on both the damage area and depth is then used to assess the condition of culverts based on a four-point condition assessment scale. The proposed framework is demonstrated with a test example. Future research would entail consolidating the concept by extensive testing and integration for real-time application.
Get full access to this article
View all access options for this article.