Waterline Detection and Water Level Estimation Based on HED Edge Detection

Due to the flooding problem which has a significant challenge towards locals who are responsible for monitoring the water level in various canals, an innovative solution for monitoring water levels is needed. Therefore, a method utilizing images from CCTV cameras for real-time water level assessment is presented. By leveraging an image processing approach including image restoration and binary image processing techniques, we utilized the HED edge detection followed by erosion to detect water lines and then estimate the water level through the comparison and matching the coordinates between detected edges and predefined water levels. Experimental results, compared to traditional edge detection like Canny and deep learning-based HED edge detection, proved the effectiveness of our proposed method in various challenging canal scenarios, exceeding the accuracy of the other two methods.