Integrating Color and Texture Features for Automated Plant Health Classification Using K-Means Clustering

In modern agriculture, the key to responding in a timely manner is the accurate judgement of plant health, which directly affects crop yield and quality. Many conventional techniques are essentially one or the other of color and texture analysis wherein performing either allows only a partial look into the information available to better classify these types of data. Summary to cope with the difficulties presented in the plant health classification task, a novel hybridized feature extraction framework combining color and texture features is proposed in this work.We use the average RGB values for color features and Grey-Level Co-occurrence Matrix (GLCM) focusing on contrast and dissimilarity properties for texture features. We merge these two features into a single feature vector which will be K-meansed. This method improves the distinction between healthy and diseased plants, which retains interpretability and accuracy better than that of traditional unsupervised classification techniques.