An Advanced Calibration Method for Image Analysis in Laboratory-Scale Seawater Intrusion Problems

Image analysis is a useful tool for visualising flow through laboratory-scale aquifers but existing methods of converting image light intensity to concentration can be labour intensive and time consuming. The new approach proposed in this study utilises the Random Forest machine learning technique to build a calibration model to replace the requirement for unique calibrations of each test aquifer. Calibration images from a previous experimental study were used to train the Random Forest model and the output was compared to the results from a high resolution pixel-wise methodology. The Random Forest model provided a trade-off in accuracy with increased efficiency and reduced sensitivity to image desynchronisation when compared to the pixel-wise method. The reduced accuracy was attributed in part to non-linear lighting distribution across the sandbox, which could be corrected by orientating the backlights effectively. Time savings of around 35% were achieved for this experimental study and this is expected to increase for larger scale studies. The new calibration approach exhibits some promising features in terms of its robustness to experimental error and its ability to process efficiently large-scale experiments in a shorter time frame.