Optimizing Image Analysis Processing in Thin Transparent Aquifers: Application to Pixel Wise Regression of Salt-Water Intrusion

Image analysis has become a standard method by which saltwater intrusion (SWI) is investigated in the laboratory. While the use of complex artificial neural networks is becoming a common analysis technique to obtain concentration fields, the standard methodology utilizes a classical algorithm which applies an augmented power-law function to each grayscale pixel. The classical method is methodologically rigorous, simple to implement numerically, and empirically robust. However, the power-law procedure involves substantial costs to the experimental process in producing calibration images for every aquifer and to computer processing times due to performing pixel-wise non-linear regression. We have developed three new classical image processing methods for SWI experiments in translucent glass-bead aquifers with the goal of optimizing the experimental and data analytic processes while maintaining accuracy and utility. First, a Laurent series provides similarly good fitting to optical grayscale data, while the function’s linearity reduces computation analysis time by a factor of a thousand—from over two hours to twenty seconds. For the second method, the Beer–Lambert Law is modified to include the optical effect of the glass beads. Applying this function form to images taken through a monochromatic light filter may decrease the number of calibration images, thereby saving the experimenter several hours of calibration time per experiment. Third, color image cameras provide different pixel intensity decreases between the three spectral channels which can be combined to produce a nearly linear correlation between source data and concentration, which gives an especially robust reduction in calibration images and rapid processing times. In our presentation, we will discuss the relative advantages and limitations of each method as they relate to the requirements and configuration of the laboratory under investigation and local analytic capabilities.