A common source of information used to identify plants is leaf shape. Methods were developed and implemented to identify in situ leaves of velvetleaf (Abutilon theophrasti), ivyleaf morning glory (Ipomoae hederacea), giant foxtail (Setaria faberi), and soybean (Glycine max) seedlings at the growth stages of (1) fully expanded cotyledons, and (2) fully developed, first true leaves. Curvature was used to describe boundaries of both completely visible and partially occluded leaves. Completely visible leaves were identified by aligning the curvature function of each leaf with that of each of the models. Functions were aligned to minimize local differences in curvature. A Fourier-Mellin correlation was used to calculate scale factors for resampling curvature functions of partially occluded leaves. Partially occluded leaves were identified by aligning the resampled curvatures with each of the models. Leaves which were oriented at oblique angles with respect to the image plane were identified by aligning the curvature function with that generated from rotated model leaves. The general shape-matching procedure was not significantly affected by differences in scale or orientation between models and unknowns. However, it was an inadequate descriptor of leaves with variable leaf serrations. Aggregate boundaries of multiple leaves could not be adequately identified by this technique. Rather, these regions required prior partitioning. Homogeneous transformations of two-dimensional model leaves were partially successful in identifying leaves oriented at angles greater than 30.
No takes yet. Share an insight, caveat, or question.
Franz et al. (1991) studied this question.