Selection of shape parameters that differentiate sand grains, based on the automatic analysis of two-dimensional images

• Image analysis of the particle projection onto a plane is really objective method.
• A shape of a grain is analysed in Krumbein method—it can serve as reference method.
• 6 shape parameters were selected that best differentiated the sand grains.
• None of them allowed the direct categorization of a particle into a Krumbein class.

A grain shape analysis of sandy deposits has implications for determining the processes that affect grain shape. So far, most methods of carrying out a grain shape analysis are based on the subjective estimation of the researcher. The purpose of this study is to indicate the shape parameter/parameters that best differentiate sand grains, and to compare the results with those that have been obtained by the Krumbein method. We determined the shape parameters of sand grains (size range from 0.71 mm to 1 mm) using photos of two-dimensional images of particle projections onto a plane. The photos were taken under an optical microscope and were then subjected to an image analysis. We selected six shape parameters that best differentiate the studied sand grains, based on the criteria of: i) the monotonicity of parameter value, which changed depending on the categorization of the grains to the successive Krumbein roundness classes, and ii) the statistical significance of differences between the optical parameter values in the individual Krumbein classes. We selected three circularity parameters (θ1, θ4 and θ5) and three surface structure parameters (κ3, κ4 and κ6). None of these shape parameters allowed the direct categorization of a particle into a particular Krumbein roundness class. Nevertheless, despite the fact that an unambiguous characterization of deposits is not possible, this method can be helpful in identifying the origin of deposits. Moreover, it is possible to develop more advanced methods (e.g., based on artificial intelligence tools), which would allow an unambiguous categorization based on the determined shape parameters.