Classification of Cultivated Seedling of Orchid Using Machine Vision

Abstract

To automate the transplant of orchid seedlings, a machine vision system was developed for their classification. The boundary of a seedling image was captured with a black and white camera and approximated using a polar coordinate. An ordered set of sequences of the length between a centroid of the seedling and the boundary was expressed as a time series. Fourier analysis was used for the time series to determine the boundary shape. A successful classification result was obtained by using a projected area and Fourier coeffcients, with vector lengths as the judment criterion.