Spherical-Based Three Dimensional Cassava Root Skeleton Extraction

Searching for high yielding cassava cultivars involves the measurements of cassava storage root morphology. This can be accurately achieved by using three-dimensional root skeletons extracting from three-dimensional images of cassava storage roots. We proposed a spherical search approach to effectively extract the three-dimensional root skeletons and a point density approach to approximate the location of stem-root junction. Our techniques were applied to three-dimensional images with uniform distributed points derived from voxel craving. The skeletal points are generated with equal distances between any adjacent points and sorted by distances from the root tip. Root segmentation is performed along with the root skeletal extraction. Moreover, fragmented skeletons caused by missing parts in three-dimensional reconstruction process are reconnected in order to increase the accuracy of root trait measurements especially the root count and root lengths.