Machine vision can be a vital tool for measuring key process parameters in many agricultural production systems, including commercial nurseries and greenhouses. One important role of machine vision in these systems may be to identify plant features and properties which allow for robotic processing, automated grading, and other tasks. A technique for identifying key features of ornamental cuttings was developed which was orientation independent, provided complete classification and interconnection of the major plant parts (e.g. main stem, petioles, and leaf blades), and relied exclusively on the morphological content of a binary image. The technique segmented a plant image into objects which could then be classified according to geometric data. Results indicated the technique was effective on geranium cutting images regardless of position or orientation. A high average percentage of total image pixels (96·6%) were classified correctly in 80 sample images tested. However, sources of error for certain misclassifications revealed limitations of this technique. Branching or fork-based segmentation can leave gaps in the geometric data required for satisfactory classification. Also, plant parts which overlap and/or occlude other parts cause conflicting geometric data which can yield classification errors. The use of spectral information may be required to extend the technique and improve robustness.