Estimating drinking water turbidity using images collected by a smartphone camera

Abstract

The lack of robust water quality data in drinking water services in many low-income settings can be attributed to inadequate funding for regular monitoring using analytical equipment. Turbidity is an indicator that is relatively quick and easy to measure; however, it still requires a turbidimeter and a trained operator. This study developed an entire smartphone camera-based application to measure turbidity in drinking water, removing both the need for external equipment and skilled labour. The application was created using a convolutional neural network, able to classify water samples into eight turbidity bins ranging from 0 to 40 NTU. The turbidity of the samples was created using formazine and kaolin clay suspensions. The in-built camera of a smartphone was used to capture images of water samples with known turbidity values. This algorithm was then embedded in a smartphone application, thereby providing an easy-to-use tool for users to estimate turbidity. Specifically, the protocol for using this application was developed with the intention that it will be used in low-resource settings by laypersons. Formazine samples achieved a turbidity classification accuracy of 98.7%, while kaolin clay samples achieved 90.9% accuracy using this method, which provides an encouraging proof of concept, as justification for further testing and improvements. HIGHLIGHTS An algorithm and accompanying protocol for estimating drinking water turbidity using only a smartphone camera were developed. The protocol is designed for use in low-resource settings by laypersons.; A correct turbidity bin classification accuracy of 98.7% was obtained for formazine samples between 0 and 40 NTU in the laboratory.;

Keywords

• cnn
• drinking water
• protocol
• smartphone camera
• turbidity