The development of dicotyledon embryos is marked by several morphologically distinct stages. The description of the developmental process of somatic embryos is typically accomplished by assigning the embryos to different morphological classes based on laborious and subjective microscopic observation. Quantitative descriptions of an embryo population and its developmental kinetics have been hindered by the inability to acquire sufficient data objectively and by the lack of adequate descriptors that discriminate the relative contributions of size and various morphological features to the embryo development. We have established a pattern recognition system capable of automatically assigning individual embryos to appropriate morphological classes by image analysis techniques. This system allows the identification of almost all the particles present in an embryo culture including normal embryos, abnormal embryos and callus. The capability of classifying abnormal embryos results in a significant improvement in quantifying the embryo development process. By constructing the distribution functions of Fourier and size features of an embryo population, statistical analysis and kinetic descriptions of a developing somatic embryo population become feasible. Monitoring changes in the Fourier and size feature distribution provides an alternative perspective in elucidating the mechanisms of plant embryo development.