KNN-based image segmentation for grapevine potassium deficiency diagnosis

In crops management, monitoring the plants health is an important task that allows early detection of nutritional deficiencies, diseases, pests, etc. that can result on important economic losses. With early detection there are more possibilities to react on an appropriate way to control the problem and assure the quality of the final products. Molecular methods performed in laboratories are a way to confirm the visual inspections on crops but the results can take valuable time. This makes the visual inspection an important task that allows early detection. Computer vision systems are an alternative solution to many problems of daily life. Many methods based on image processing have been developed focused on early detection of diseases and nutritional deficiencies. Specialized computer vision systems can be a support on decision making to appropriate crop management. In this paper, an image processing method to diagnose and classify grapevine leaves with certain level of potassium deficiency is proposed. The proposed segmentation method based on K-Nearest Neighbors (KNN) were compared to methods based on histogram. KNN proved to have better results specially when the environment were images are acquired is less controlled.