Hybrid IoT-VIoT System for Real-Time Water-Level Monitoring Using Computer Vision

Efficient water resource management is critically important for arid regions such as southern Kazakhstan. This paper presents a hybrid Internet of Things (IoT) and Vision-based Internet of Things (VIoT) architecture for real-time monitoring of water levels in irrigation channels. The proposed system integrates an ultrasonic water-level sensor, an IP camera with edge-based computer vision processing on a Raspberry Pi, wireless communication, an autonomous solar power supply, and discharge estimation using Manning’s equation. The VIoT subsystem applies image processing techniques, including gauge calibration, Canny edge detection, and pixel-to-metric conversion, to automatically estimate water level from captured video frames. Water-level measurements obtained from IoT sensors and video-based analysis are combined through synchronised data fusion to improve monitoring accuracy and reliability. The hybrid approach leverages the complementary strengths of IoT and VIoT by combining continuous quantitative sensing with visual verification capabilities. Field experiments conducted on the Merke River in the Zhambyl region of Kazakhstan over a 14-day observation period demonstrated stable real-time operation with RMSE = 0.311 cm, MAE = 0.279 cm, and Pearson r = 0.99 between the ultrasonic sensor and the vision-based estimates. Sensitivity analysis indicated that water level is the most influential parameter in Manning-based discharge estimation, confirming the importance of accurate level detection. The proposed system improves reliability by cross-checking independent data sources, making it applicable to monitoring water levels in agricultural regions.