Journal of Animal Science, Vol 69, Issue 9 3807-3816, Copyright © 1991 by American Society of Animal Science

JOURNAL ARTICLE

Image analysis and nonlinear modeling to determine dimensions of wet-sieved, masticated forage particles

J. M. Luginbuhl, D. S. Fisher, K. R. Pond and J. C. Burns
Dept. of Anim. Sci., North Carolina State University, Raleigh 27695-7621.

A procedure to estimate dimensions of digesta particles was developed and evaluated. After wet sieving, particles of masticated switchgrass and bermudagrass hays retained on sieves with apertures of 1.0 mm2 or larger were placed on moist white filter paper, separated manually, and photographed. Photographs were scanned with a flat-bed scanner, scaled to size, and digitized. Projected area, perimeter, length, and width of particles were determined by a microcomputer system. Variables derived for each particle included the form factor, which was defined as (4 x pi x area)/perimeter2, and each particle's length: width ratio. Ninety-three percent of data from two trials with steers had significant lack of fit (P less than .05 to P less than .0001) to normal, lognormal, Weibull, or gamma distributions. As an alternative, individual particle measurements were summed on a cumulative percentage basis for each variable and were fit to an inverted Gompertz function to estimate median, mode, and mean dimensions of scanned particles. Estimates from the equation fit the distribution well; asymptotic SE averaged 3.5 and 1.4% of parameter means for dimensions of masticated switchgrass and bermudagrass hay particles, respectively. No analytic solution exists for the mean, which must be estimated numerically, but analytic solutions are available for the median and mode. Use of this equation will prevent bias caused by lack of fit to a particular distribution and yield more accurate estimates of mean particle dimensions than arithmetic means.