Characterization of growing lettuce from density contours—I. Head selection

摘要

Image processing techniques have been applied to the (one-dimensional) density profiles one obtains when rows of growing iceberg lettuce are scanned with a moving X-ray source. Separation of the X-ray scan into individual head clusters is achieved by a valley seeking algorithm which depends, in turn, on a fast convex hull algorithm, which is described. Remaining clustering errors are recognized by n-space linear partitioning, resulting in an overall error rate of about 2%. Such levels of accuracy are required in the present applications which involve the breeding of experimental plant varieties and the management of commercial fields. This is one in a series of papers using moving scanners to characterize standing row crops.