A Using the Canny Method with Deep Learning for Detect and Predict River Water Level

Abstract

Recent years have witnessed a significant rise in global river water levels, driven by heavy rainfall events linked to climate change, resulting in severe flood incidents and highlighting the need for effective mitigation strategies. To address this critical issue, this article introduces a detection and prediction system for monitoring rising river water levels utilizing advanced computer vision techniques. The system is based on deep learning models are VGG16and 2D convolutional neural networks (CNNs) that are trained on river image datasets. It starts with heavy image preprocessing are normalization, resizing, Canny Edge Detection to improve edge quality. This is used to produce very exact water level measurements. The models use transfer learning and hyperparameter tuning over a wide grid to maximize monitoring accuracy over a large number of river conditions. The system performance is then thoroughly evaluated by performing extensive testing, and the VGG16 model is proven to exhibit high classification ability with the overall accuracy and precision scores no less than 98.3 % of all classes of river water level. The performance of the CNN model reaches 99.99% on testing data as well, which emphasizes its stability in real-world conditions. This novel combination of deep learning and optimized edge detection algorithms represents a powerful new capability for water level detection and flood control.