Plant Identification From Leaves Using Quasi-Sensor Fusion

Abstract

The method described here identifies plants by using a machine vision technique. This method achieves effective image detection independent of surrounding conditions, dimensionless image detection in each growth stage, and determination of the critical factor for discriminating individual plants. These are the fundamental factors for successful automatic thinning, cropping, weeding, and harvesting using intelligent agricultural robots. Color, aspect ratio, size, radius permutation in leaf profiles, complexity, and curvature are used to classify each plant. Effective discrimination is obtained by using a quasi-sensor fusion combined with a total occurrence range for decision making.