It is a long-lasting question whether we can determine the concentration and particle size by only looking at the color of turbid water. We used an emerging technology called hyperspectral imaging to answer the questions. This new imaging technology collects and processes information along the electromagnetic spectrum, creating high-resolution color signatures for each pixel (e.g. intensity for every 5 nm of the wavelength) which can then be used to identify objects and processes in an image. Our experiments will explore if the hyperspectral signature of sediment mixtures against plain water samples can suggest the concentration and grain sizes of the suspended sediment. We performed a series of experiments with different particle sizes and concentrations in the lab. A clustering model was built to identify the pixels of suspended sediment imaging based on the spectrum similarity as well as extracting spectral graphs to contrast intensity reflections. The preliminary results show that the color of the water sample can inform the particle size and concentration. We will present the final result at the conference. The experiment and the derived machine learning models can be used to inform remote sensing using satellites to monitor sediment transport on a large scale. The developed technology can be further extended to study algal bloom and other water quality issues.
.jpg "Ruo-Qian Wang photo")