AUTOMATED CLASSIFICATION OF SOMATIC PLANT EMBRYOS

A computer vision system for the classification and selection of somatic plant embryos was developed. The proportions of embryos at different developmental stages in liquid samples from the bioreactor give an estimate of the state of the embryo production process. The considerable shape variation of the embryos and the presence of undifferentiated cell aggregates make the automated classification difficult. However, the specific classification algorithm developed for birch (Betula pendula Roth.) somatic embryos showed that reliable automated recognition of embryos is possible without a significant number of false selections. The system constructed can also be used to automatically monitor cell suspension cultures of other plants. However, a specific recognition algorithm needs to be implemented for every plant species. Algorithms already exist for the recognition of birch and Norway spruce (Picea abies) somatic embryos.