Methodology based on machine learning and digital image processing for estimating turbidity in the water purification process

Turbidity is one of the physical parameters used in the study of water quality. The traditional methodology used to measure turbidity involves the use of a turbidimeter, which uses optical theory. It studies how water particles scatter light, resulting in a decrease in the intensity of the light transmitted through a sample. Turbidity measurements indicate the "Total Nephelometric Units (NTU)," the unit of measurement corresponding to the parameter. Different measurement approaches have been used, such as the aforementioned optical approach and chemical methods. These methods require specialized personnel and equipment, and the equipment is expensive. To address this problem, one alternative is the use of machine learning algorithms; these greatly reduce the cost of manual methodologies. The algorithms allow for predictions of turbidity values, as would be possible with an industrial turbidimeter. Motivated by the above, we propose the implementation and evaluation of key machine learning models that predict different turbidity values.