Image analysis with artificial neurons to select cutting planes for the micropropagation of syngoniums

• Published in: Computers and electronics in agriculture Vol. 10; no. 3; pp. 189 - 202
• Main Authors: Davis, P.F., Tillett, R.D.
• Format: Journal Article
• Language: English
• Published: Amsterdam Elsevier B.V 01.06.1994; Elsevier
• Subjects: Agronomy. Soil science and plant productions; ARACEAE; AUDIOVISUAL AIDS; AUTOMATION; AUTOMATISATION; AUTOMATIZACION; Biological and medical sciences; BOUTURE; CUTTINGS; ESQUEJES; Fundamental and applied biological sciences. Psychology; General agronomy. Plant production; MEDIOS AUDIOVISUALES; MICROPROPAGACION; MICROPROPAGATION; MOYEN AUDIOVISUEL; Vegetative propagation. Micropropagation; Vegetative propagation. Sowing and planting. Harvesting; Monocotyledones; Automatic selection; Tissue culture; Ornamental crop; Micropropagation; Cutting plane method; Microplant; Discriminant function; Image analysis; Artificial neuron; Araceae; Indoor plant; Angle; Angiospermae; Computerized processing; Spermatophyta; Artificial intelligence; Computer aid
• ISSN: 0168-1699; 1872-7107
• DOI: 10.1016/0168-1699(94)90040-X

This report describes an investigation of the use of image analysis for automatically selecting the cutting planes for syngonium microplants. An artificial neuron was trained to select a cutting line across each image viewed, and another neuron was trained to compare viewing angles. Each syngonium plant studied was held upright and rotated about a vertical axis while it was viewed by a video camera from the side. At each of a set of viewing angles, a tentative cutting line was selected between clumps of shoots. The position of the tentative cutting line and other features were then used to automatically evaluate the viewing angle until the best view for selecting a cutting plane had been found. The cutting plane was taken as the plane containing the viewing direction and cutting line. In trials with 101 plant images, the success rate in automatically selecting a suitable cutting line at a given viewing angle was 92% and the overall success rate, when selecting both line and angle, was 80% (including 6% borderline cases).