Smart Valve Detection System for Water Distribution Networks

Abstract

The water distribution network is one of the fundamental and expensive infrastructures to sustain the urban life. The aging of these infrastructures and pipe deterioration are becoming major issues to tackle as it leads to massive water loss and environmental adversities. To combat the aforementioned issues, the water municipalities have included condition assessment programs to assess the internal condition of the pipelines. The assessment is usually carried out through the in-pipe inspection device with closed-circuit television (CCTV) system to videotape inside the pipelines. However, the in-pipe inspection device faces challenges to navigate through the butterfly valves inside the pipelines. This impedes the videotaping process and disrupts the condition assessment process as well. Hence, this paper proposes a smart valve detection system to detect valves in real-time by adopting NASNet architecture combined with a Faster R-CNN object detector. The experimental results from the proposed system show that the integration of valve detection into the in-pipe inspection tool can help the device to enable the control mechanism and navigate through the butterfly valves and also, aid in the efficient management of the water infrastructure.