AUTOMATIC PLANT FEATURE IDENTIFICATION ON GERANIUM CUTTINGS USING MACHINE VISION

Abstract

ABSTRACT A machine vision technique was developed which analyzes a two-dimensional binary image of a singulated geranium cutting and identifies the branching stem structure, including main stem and petioles. The analysis technique was based on creation of a directed graph data structure which contains information required to rapidly perform plant part identification. Size, shape, and location data were utilized to classify objects as particular plant features. Evaluation of the image analysis technique indicated good characterization of the binary structure of geranium cuttings in a timely manner as required for use in a robotic workcell. Identification of the main stem, petioles, growth tip, and geometry of the interconnections of the plant parts was successfully performed. Overlapping sections (e.g., petiole crossings) and occlusions (e.g., leaves over stem segments) contributed to identification errors.