Color Segmentation and Machine Learning for Disease Recognition in Rice Crop Leaves

Timely detection of diseases in various crops is a necessary task to ensure sufficient production of food sources. Visual analysis by an expert is the method traditionally used for this activity, so it is subjective and prone to errors. In this paper, we propose a color segmentation method and a feature analysis for the recognition of rice crop leaves using machine learning. We use balanced sets of images and propose a set of experiments that allow us to discover the features that influence the classification indices, like the need to identify more precise characteristics for the classes of similar leaves or the disease regions. Results show that some features of texture and color are irrelevant for disease recognition.