NTT DOCOMO group to release trial AI system for monitoring river and reservoir water levels

Japan’s telecommunications company NTT DOCOMO, NTT Communications and NTT Comware announced that they will provide a trial version of an artificial intelligence (AI)-based system for monitoring water levels of rivers and reservoirs nationwide. Through the trial, the companies are to improve the system’s AI functions and usability, aiming at a commercial launch by March 2024.

The system is said to eliminate the need for physical water-level gauges. Instead, it uses virtual gauges and AI technology to visualise real-time images and time series data of water levels on a screen. The system can automatically notify authorities if a predetermined water level is exceeded.

The system uses cameras installed near a river or reservoir to capture video images. A virtual gauge is overlaid on the video and AI video-recognition technology determines the water level in real time. Specifically, the system uses an algorithm and AI image-segmentation technology to analyse the ratio of the water surface to the land area displayed on the virtual gauge. Accuracy can be improved by using multiple virtual gauges.

Personnel monitoring rivers and reservoirs for municipalities can check water levels and configure AI settings on a management screen in the cloud. The system can be livestreamed so local residents are informed. The system incorporates two other key technologies, EDGEMATRIX and SmartMainTech.


NTT Com video analytics platform processes high-volume video data in an Edge AI Box, enabling rapid video analysis, water-level determination and streaming.


DOCOMO — as part of its SmartMainTech series of solutions for infrastructure maintenance — determines water levels by combining DOCOMO AI technology and Edge AI Box to analyse videos of rivers and reservoirs, and link results to the SmartMainTech cloud. The solution is said to incorporate industry-specific AI, data analysis and visualisation technologies.

The system features the use of low-cost virtual water-level gauges which make it practical to monitor and maintain compared to real gauges. They cannot be easily damaged by floods and multiple virtual gauges are installed for each camera to increase system accuracy. Next, the use of real-time video helps confirm rapid water level changes along with batch video and data monitoring which allow the viewing of real-time videos and water level graphs on the same management screen. Notifications will be sent automatically via email or text if water level exceeds. Further, videos can be easily linked to video-distribution sites to keep nearby residents up-to-date and Edge AI Box performs video encoding in the field, enabling real-time video delivery.

The three companies each performed a specific role in the development of the system: (1) DOCOMO developed the AI used in SmartMainTech; (2) NTT Com provided EDGEMATRIX service, cameras and Edge AI Box; and (3) NTT Comware developed AI applications under the SmartMainTech brand.

In recent years, river floods caused by typhoons and torrential rains have become more severe due to climate change, increasing the need to ascertain water levels and take countermeasures quickly. But the use of conventional physical water-level gauges is costly and labour-intensive in terms of installation and maintenance. The NTT DOCOMO group’s low-cost and low-maintenance system is expected to expand real-time monitoring of large bodies of water, small rivers and reservoirs that have the potential to cause disasters.