Automated detection of anomalies in sewer closed circuit television videos using proportional data modeling

Abstract

Sewer pipeline condition information is usually collected using closed circuit television (CCTV). Moreover, in order to evaluate the condition of pipeline, data should be processed by a certified operator, which is time consuming, costly, and error prone due to operator's skillfulness or fatigue. Automating the detection of anomalies can reduce time and cost of inspection while ensuring the accuracy and quality of assessment. However, considering various types of defects in sewer pipelines and numerous patterns of each, it seems to be difficult to detect the defects using computer vision techniques. This paper presents an efficient anomaly detection algorithm to support automated detection of sewer defects from data obtained from CCTV inspection videos. In this model Hidden Markov Model (HMM) for proportional data modeling is employed theoretically and its performance of anomaly detection in an example of sewer CCTV videos has been assessed. The algorithm consists of modeling conditions considered as normal and detecting outliers to this model.