Machine learning based smartphone spectrometer for harmful dyes detection in water

摘要

The development of low cost and portable instrumentations to monitor water quality without access to sophisticated laboratory tools has attracted an increasing amount of attention recently. In this study, we design a portable, cost-effective, field-deployable, plastic fiber based smartphone spectrometer system to detect harmful dyes in water. The proposed system uses a custom cradle to transmit built-in flash light via fiber cable, and to capture the visible absorbance spectroscopy by camera of smartphone. The captured images were trained and tested with machine learning algorithms using color features. Experimental investigations for detection of methylene blue (MB) absorbance in water shows that the proposed system can detect the MB with % 100 accuracy. With advanced microfabrication and data processing tools, our proposed system will be more compact, lighter, cost-efficient and will be enhanced with more features for the next generation of sustainable water management applications.