Identification and Segmentation of Occluding Groups of Grain Kernels in a Grain Sample Image

Algorithms were developed to solve the problem of identification and segmentation of occluding groups of grain kernels in a grain sample image. The first algorithm characterized each object in a binary image of a grain sample as either an isolated kernel or a group of occluding kernels by determining the degree of overlap between each object in the input image and its inertial equivalent ellipse. If the degree of overlap was significant, the algorithm characterized the object as an isolated kernel. Otherwise, the algorithm marked the object as a group of touching kernels to be processed by the second algorithm. The second algorithm separated individual grain kernels in binary images of touching kernels. Segmentation lines between nodal points (points where the individual kernel boundaries intersect) were drawn by the algorithm. Nodal points were determined by evaluating the curvature along the boundary and selecting those points at which the curvature fell below a threshold. In situations where more than two nodal points were found, a `nearest-neighbour criterionʹ was used to draw the segmentation lines. The algorithm performed with 99% reliability on images containing touching kernels of barley, hard red spring (HRS) wheat, and rye. The reliability was considerably lower for images containing kernels with rough boundaries.