Publication: Smartphone based imaging with spirodella polyrhiza for water quality monitoring with the aid of computational modelling.
No Thumbnail Available
Date
2025-06
Authors
Tan, Win Hung
Journal Title
Journal ISSN
Volume Title
Publisher
Abstract
The demand on clean water had stimulated the development of various water monitoring technique. From the conventional to the advance monitoring, the efforts were directed towards developing rapid, sensitive and easily operated methods. To address the problem whereby the conventional water quality monitoring relied on the chemical usage and tedious procedure, the present study proposed an innovative smartphone-based imaging technique for the detection and prediction of water contamination. The prediction is based mainly by interpreting the physical changes of a bioindicator, after exposed to different type and concentration of contaminants. The research initiated by evaluating effects of macronutrients on the biochemical changes in a species of duckweed, Spirodella. polyrhiza with machine learning approaches, followed by establishing the framework for smartphone-based imaging estimation of mass, chlorophyll and anthocyanins. A smartphone was used to capture the images of S. polyrhiza and captured image was processed with ImageJ software to extract the color signal. The color signal from the images were correlated with the exact determined amount of chlorophyll and anthocyanins. The mean green (G) value of the image was a robust parameter to predict chlorophyll contents in S. polyrhiza with a high r2 of 0.9693 and the lowest error compared to other color parameters and their combinations. Furthermore, the accuracy for the estimation of mass and chlorophyll under magnification was tested. A similar approach was carried out and the magnifications were employed with a microscope. The image captured was analyze and correlation was established. For chlorophyll estimation, the best magnification to be used was 2.5x magnification, and the derived color combinations, square root of R+G+B has the highest r2 value which is 0.7166. Machine learning algorithms were used to further improved the generated model to the r2 value of 0.72. External factors were evaluated by capturing images under the influence of distance, light intensity, correlated color temperature and smartphone model variation. The external factors were found to affect the output from images. Final study involved the development of classification model for estimating type of contaminants and prediction model for anticipate the concentration of contaminants. The classification models were good in classify the type of dye and predicts with general class of contaminants. The predictive model developed works well with the concentration of oil, methylene blue, chromium, cadmium and silver (r2= 0.8561, 0.9464, 0.8326, 0.9727 and 0.9529 respectively). Smartphone-based imaging with S. polyrhiza had been proven workable in predicting the contamination in water.