A Vision-Based System for Real-Time Monitoring of Water Quality in Natural and Artificial Reservoirs

Manual sampling and laboratory testing are the mainstays of traditional water quality monitoring methods, which causes delays in data processing and decision-making. This study proposes an enhanced monitoring of water quality system that uses machine learning algorithms to classify and segment water samples in real-time, thereby overcoming these restrictions. Water samples are classified as “Clean” or “Polluted” having an accuracy of 83.99% using a CNN model. The CNN model, trained over 20 epochs, effectively learns patterns from the dataset to deliver precise results. Additionally, a U-Net model is implemented for image segmentation, generating masks that accurately highlight areas of interest within the water samples, with a segmentation accuracy of 68%. Streamlit's intuitive interface makes it simple for users to upload images and get immediate feedback on water quality. This system helps to improve the health of the environment monitoring with automating the method of water quality evaluation. It does this by providing a simpler, quicker, and scalable solution that enables prompt, data-driven actions.