Innovación en Minería de Datos para el Tratamiento de Imágenes: Agrupamiento K-media para Conjuntos de Datos de Forma Alargada y su Aplicación en la Agroindustria

Abstract

This paper presents an innovative modified method of K-means clustering based on the set theory together with its application in the processing images of the agroindustry field. Traditional K-means permits the clustering of sets in subsets by means of defining their center according to the distance formula. When the data is concentrated in forms without a hyper-spherical sense, this tool allows the center of the set, with a single point, to become a subset of many points. In this article we present a modification of the distance formula that allows giving more flexibility for the study of cases in agriculture. Using numerical examples, the functionality and applicability of the modified method of K-means grouping is evaluated in infrared images from water deficit tests in wheat.