Organization: Mitsubishi Electric Engineering
Co-Developer(s): Mitsubishi Electric Research Laboratories, Mitsubishi Electric
Year: 2018

Because water levels change frequently, government agencies are continuously monitoring and recording major rivers in order to manage construction plans, flood control, to secure the water supply, and to prevent natural disasters. Currently, water levels are commonly measured by automatic recording devices such as those that use hydraulic pressure methods—where the pressure sensor installed on the river bed sensors the water pressure changes—or by floating devices, which sense up and down movement. However these types of sensors are easily broken or lost due to the water pressure that occurs during heavy rain or flooding and valuable data is lost. Deep Learning Based Water Gauge (Powered by Maisart) is a deep-learning based water gauge which provides non-contact water level measurement for rivers by combining water region extraction from camera images and 3D measurement data obtained by a 3D laser measurement unit. Compared to conventional products of other companies, the stability of measurement, and the adaptability and accuracy under extreme weather and sunshine conditions, are greatly improved by adopting the deep learning technology to extract water region from images.

For more info: