Selection of descriptors for particle shape characterization

Conventional shape descriptors, formed from a ratio of two particle size measurements, are among the simplest of the many methods used for quantitative particle shape characterization. However, a significant limitation of using one of these shape descriptors is that its value is often not unique to a specific shape. Use of several different shape descriptors may circumvent this problem but, as particle size can be defined in a large number of ways, a similarly large number of shape descriptors can be defined. While some differ substantially, others are only subtly different, conveying similar information. Thus, it is not obvious which of the many possible descriptors should be utilized. In this paper, two-dimensional particle shape descriptors obtained by image analysis of six different commercially sourced powders were considered. Techniques of cluster and correlation analysis were applied to assist in identifying redundant descriptors for shape characterization of these powder particles, allowing for efficient description of shape using a reduced set. It was found that at least two descriptors are required: aspect ratio and the square root of form factor. Significantly, each descriptor is most sensitive to a different attribute of shape: elongation and ruggedness, respectively.