[A-TT30] Machine Learning Techniques in Weather, Climate, Ocean, Hydrology and Disease Predictions

[ATT30-P02] A Study on the Establishment of Open Channel Water Level Recognition Method Using Artificial Intelligence Model in the Absence of Historical Water Level Imagery

*JIAN-MING Wang¹, Po-Tsang Chen² (1.Postgraduate, Department of Water Resources Engineering and Conservation, Feng Chia University, Taiwan (R.O.C.), 2.Assistant Professor, Department of Water Resources Engineering and Conservation, Feng Chia University, Taiwan (R.O.C.))

Keywords:Artificial Intelligence, Convolutional Neural Network, Water Level Recognition, Intelligent Water Gauges

Implementing artificial intelligence models for pattern recognition commonly requires preliminary training with historical data of the subject in question. Specifically, for the application of recognizing water levels in channels, it is imperative to possess a database of varied water heights at the designated locations. This study is intended to evaluate the practicality of employing existing CCTV systems to capture and utilize imagery of channels for recording water levels, in the absence of historical elevation water level imagery. Utilizing images of channels at low water levels during the dry season, this research involves artificially generating images that represent various water heights for AI model training. After the training phase, the model is tested with actual images of high water levels, and the outcomes of these tests are used to formulate standard operating procedures. Repurposing existing CCTV into intelligent water gauges not only substantially reduces the installation costs of water level sensors across catchment areas but also provides a fuller understanding of catchment runoff information. This improvement is invaluable for assessing the rainfall-runoff process during disaster reduction phases and enhances the capability to manage disaster situations within a jurisdiction effectively.