Minimum image quality for reliable optical characterizations of soil particle shapes

Abstract

Remarkable advances have been seen in image-based methods for automating soil particle shape characterizations in the last decade. However, the accuracy and reliability of image-based methods has rarely been questioned. This study shows that image quality affects the computational results of particle shape descriptors, including aspect ratio, sphericity, convexity, circularity, and roundness. These descriptors display a hierarchy of resistance to the effects of low image quality. The particle length, perimeter, and area are used as controlling parameters for quantifying the influence of image quality. The minimum requirements for ensuring reliable image-based shape characterization of these parameters are established.