DBPapers
DOI: 10.5593/SGEM2014/B23/S10.008

AUTOMATIC TEMPLATE GENERATION FOR INDIVIDUAL TREE IDENTIFICATION IN REMOTE SENSING DATA USING (1+1) EVOLUTIONARY ALGORITHM

L. Gulbe
Wednesday 1 October 2014 by Libadmin2014

References: 14th International Multidisciplinary Scientific GeoConference SGEM 2014, www.sgem.org, SGEM2014 Conference Proceedings, ISBN 978-619-7105-12-4 / ISSN 1314-2704, June 19-25, 2014, Book 2, Vol. 3, 63-70 pp

ABSTRACT
Accuracy and computational speed of the template matching procedure for individual tree identification depends on the choice of templates representing tree crowns. The aim of this paper is to present a novel approach for optimal template generation using an evolutionary algorithm. Suggested methodology enables automatic generation of the templates by minimizing the commission and omission errors of tree identification according to the reference set with selected tree crown areas. Template matching procedure with automatically generated templates for the study site achieved 74% accuracy (205 of 277 trees) with 27 false positive records, while manually selected templates from the same image resulted in 53 % accuracy (147 of 277 trees) and 25 false positive records.

Keywords: template matching, automatic template generation, (1+1) evolutionary algorithm, individual tree identification, high resolution remote sensing data