Unsupervised classification of remote sensing images combining Self Organizing Maps and segmentation techniques

• Author(s): Galindo Gonzalez, Diana Rocío
• Date: 2013
• Persistent ID: http://hdl.handle.net/10362/9186
• Origin: Repositório Institucional da UNL
• Subject(s): Classification; Remote sensing; Self-Organizing maps; Visualization technique
• Document Type: Master thesis
• Language: English
• Advisor(s): Pebesma, Edzer; Henriques, Roberto André Pereira; Pla Bañón, Filiberto
• Contributor(s): RUN

Description

Dissertation submitted in partial fulfillment of the requirements for the Degree of Master of Science in Geospatial Technologies.

This study aimed a procedure of unsupervised classification for remote sensing images based on a combination of Self-Organizing maps (SOM) and segmentation. The integration is conceived first obtaining clusters of the spectral behavior of the satellite image using Self-Organizing Maps. As visualization technique for the SOM is used the U-matrix. Subsequently is used seeded region growing segmentation technique to obtain a delimitation of the clusters in the data. Finally, from the regions of neurons in the U-matrix are deduced the clusters in the original pixels of the image. To evaluate the proposed methodology it was considered a subset of a satellite image as use case. The results were measured through accuracy assessment of the case and comparing definition of the obtained clusters against each technique separately. Cramers'V was used to evaluate the association between clustering obtained each method separately and reference data for the specific use case.