Identification of high-injury-risk glass contaminants using simple shape measures of fragment outlines

Horticultural peat supplied to the domestic market is mainly hand manipulated by gardeners. For a glass contaminant to cause injury it must be angular (or pointed) and must have a significant protrusion at its angular end. These features are easy to identify visually, and shape measures specific to the determination of high-injury-risk do not exist. This study aimed to identify the best simple shape measure for the estimation of risk of injury from glass contaminants that may be found in bags of horticultural peat. Glass contaminants included in peat were detected using dual-energy X-ray absorptiometry (DXA), and ranking of various shapes according to risk of injury was carried out visually, as well as using simple shape measures. The simple shape measures used were compactness, eccentricity, elongation, radius of gyration, principal axes ratio, bounding rectangle to outline area ratio, object-ellipse major-to-minor axes ratio, convexity, convex hull area to object area ratio, and circular variance. Comparison of each of the rankings based on a simple shape measure with the visual ranking was carried out using Kendall's rank correlation test. Compactness correlated best (τ = 0.735, p < 0.000005) with the lowest computational demand followed by object area to convex hull area ratio (τ = 0.694, p < 0.000005), principal axes ratio (τ = 0.563, p = 0.000002) and bounding rectangle area to object area ratio (τ = −0.498, p = 0.000025). The use of compactness as a means of identification of risk of injury is feasible for online application on each of three orthogonal radiographic projections of peat volume.