Smartphone-Based Turbidity Estimation with Inherent Calibration

Water is one of the most vital natural resources; it is essential to daily life. In many developing countries, however, access to clean and safe water is a crucial issue due to a lack of economic and infrastructure resources. The most broad and universal measure of water quality is turbidity. Traditional lab-use turbidimeters, though highly accurate, may be prohibitively expensive for wide-scale use. Thus, proposing cost-effective water turbidity estimation systems has long been a trend. In this paper, we introduce an innovative water turbidity estimation system. Different from other approaches which have only one water sample for analysis, our system consists of a low-cost illuminated cuvette holder containing a test and a control sample, where the control always contains clear water. Acquired images of the two samples are input into an image processing chain to estimate the sample's turbidity. We evaluated our approach in both lab and in-situ environments and found that in the lab environment our approach achieved mean error of 19.07 NTU over the range 16.1 to 417 NTU compared to 69.86 NTU without using the control sample. In outdoor real-world use, we found mean error of 20.68 NTU in the shade and 57.34 NTU in full sun.