Remote sensing image classification by cluster labeling using stochastic distances

• Grant number: 16/06242-4
• Support type: Scholarships in Brazil - Scientific Initiation
• Effective date (Start): June 01, 2016
• Effective date (End): May 31, 2017
• Field of knowledge: Physical Sciences and Mathematics - Geosciences
• Principal Investigator: Rogério Galante Negri
• Grantee: Gabriela Ribeiro Sapucci
• Home Institution: Instituto de Ciência e Tecnologia (ICT). Universidade Estadual Paulista (UNESP). Campus de São José dos Campos. São José dos Campos , SP, Brazil

Abstract

Remote sensing image classification is one of the most important applications of Pattern Recognition in environmental studies. Generally image classification methods have supervised or non-supervised learning. While supervised learning methods perform classification by means of a function or decision rule modeled using a priori information, the non-supervised methods models the knowledge through analogies observed from the data. However, both methods have limitations. The quality of the supervised classification results is directly related to the quality of the training patterns set provided a priori, which cannot always be guaranteed. Regarding the non-supervised classification, it is necessary to interpret the relationship between the identified clusters and the different classes in the problem, which can be a complex task. An alternative to dealing with the weaknesses of both paradigms is offered by the semi supervised learning, whose motivation is to combine concepts of supervised and unsupervised learnings. In this context, this research project proposes the formalization and implementation of semi-supervised classification method that combines classic tools of the Pattern Recognition area: the Hierarchical Divisive Algorithm, K-Means and Stochastic Distances. From a set of clusters defined by combination of the Hierarchical Divisive Algorithm and K-Means, in an non-supervised way, the Stochastic Distances are used for labeling each of these groups. Case studies on the classification of land use and land cover in an Amazonian region will be conducted in order to compare the proposed method with other classification methods known in the literature. (AU)

Scientific publications

(References retrieved automatically from Web of Science and SciELO through information on FAPESP grants and their corresponding numbers as mentioned in the publications by the authors)

SAPUCCI, GABRIELA RIBEIRO; NEGRI, ROGERIO GALANTE. Hierarchical clustering and stochastic distance for indirect semi-supervised remote sensing image classification. SN APPLIED SCIENCES, v. 1, n. 3 MAR 2019. Web of Science Citations: 0.