Automatic measurement of river water-level using image-based computer vision

Abstract

Monitoring the water reservoir in hydroelectric plants is of fundamental importance for the manage-ment of energy production, flood warnings and planning of water resources. To that end, telemetricand electronic meters allow water levels to be automatically measured and monitored in the hydroelec-tric power plant. Such methods may have flaws or inaccuracies, necessitating a second measurementmethod. Video surveillance has come to be widely used for monitoring as a redundant system, butit is still subject to human errors in visually reading information. In this context, this work proposesa redundant automatic measurement method using computer vision. As a case study, images fromconventional cameras from the Jirau Hydroelectric Power Plant, on the Madeira River, Brazil, wereused. The results obtained show RMSE, MAE and푅2errors of 0.045, 0.0328 and 0.9946 respectively.Such results show that the proposed model can collaborate as a redundant monitoring method