IDENTIFICATION OF IDEALIZED LEAF TYPES USING SIMPLE DIMENSIONLESS SHAPE FACTORS BY IMAGE-ANALYSIS

Identification of plants is important to develop robotics for pest, di sease, and weed control with machine vision. Leaf shape is a common so urce of information used to identify plants. Intelligent vision system s are the next generation in machine vision. The addition of intellige nce into vision systems requires an understanding and structuring of h uman visual techniques. Image acquisition and processing software was developed and evaluated to identify idealized leaf types using simple dimensionless shape factors. Compactness, roundness, elongation, lobat ion, and roughness were defined and used as the shape factors. The sha pe factors require neither a heavier computer load nor a larger amount of computer memory. They are not only size and rotational independent , but also suitable for imitating human visual memories. The Makino's 50 figures which illustrated the terminologies applied to the lamina s hapes of leaves by a diagram were selected as idealized leaf types. It was theoretically possible to identify 1,280 leaf types using the sha pe factors by image analysis with sufficient resolution. Results indic ate the simple dimensionless shape factors will be useful to identify plants by their leaf shapes, and to compose knowledge bases of the int elligent vision systems.