Weeds and Plants Recognition using Fuzzy Clustering and Fractal Dimension Methods for Automatic Weed Control

Abstract

Destructive impacts of herbicide usage on environment and water contamination have led researcher orientation toward finding solutions for their accurate use. If density and weeds species could be correctly detected, patch spraying or spot spraying can effectively reduce herbicide usage. A machine vision with precise automatic weed control system could also reduce the usage of chemicals. Machine vision is a useful method for segmentation of different objects in agricultural applications, especially pattern recognition methods. Many indices have been investigated by researchers to perform weed segmentation based on color information of the images. The purpose of this research is to develop machine vision system that can detect weeds density and variability. In this study the relation between three main color components (red, green & blue) of the images and color feature extraction (hue, saturation, intensity) were used to define weeds density. Fractal dimension was used as the method to define shape features to distinguish different type of weeds and plants. Weeds and plants were segmented from background by obtaining hue value and its shape was obtained by fractal dimension value. The results showed that fractal dimension value for each plant species has specific differences. Corn plants have fractal dimension values in the range 1.148 to 1.268, peanut plants have fractal dimension values in the range 1.511 to 1.629, while the weeds have fractal dimension values in the range 1.325 to 1.497.