A New Color-based Lawn Weed Detection Method and Its Integration with Texture-based Methods: A Hybrid Approach

Abstract

We propose a color-based weed detection method specifically designed for detecting lawn weeds in winter. The proposed method exploits fuzzy logic to make inference from color information. Genetic algorithm is adopted to search for the optimal combination of color information, fuzzy membership functions, as well as fuzzy rules used in the method. Experimental results show that the proposed color-based method outperforms the conventional texture-based methods when testing with a winter dataset. In addition, we propose a hybrid system that incorporates both texture-based and color-based weed detection methods. It can automatically select a better method to perform weed detection, depending on an input image. The results show that the use of the hybrid system can significantly improve weed control performances for the overall datasets.