Gradient Field Distribution and Grey Level Co-occurrence Matrix techniques for automatic weed classification

Nowadays it is a requirement to adopt a greener approach in the management of plantation especially when dealing with chemical herbicide to control weed infestation. To do so, selective patch spraying method is a necessity since it can help in minimizing the volume of usage of the herbicide. As such, an intelligent system that can differentiate the different weed is desirable. In this work, we have adopted an image processing approach to detect and classify weed according to its class, namely as either broad or narrow, such that the selective patch spraying strategy can be implemented. This paper describes the procedures involved and its main focus is on the combined use of Gradient Field Distribution (GFD) and Grey Level Co-occurrence Matrix (GLCM) algorithms to extract new feature vector set. The results obtained suggest that the new feature vectors, derived from the GFD and GLCM techniques combined, has unique characteristics that enable perfect discrimination between the two types of weed. Thus, perfect classification was possible when tested with 400 samples of weed images comprising of both types of weed.