Methods for fuzzy classification and accuracy assessment of historical aerial photographs for vegetation change analyses. Part I: Algorithm development

F. Okeke, A. Karnieli

Research output: Contribution to journalArticlepeer-review

45 Scopus citations

Abstract

Image classification of historical aerial photographs is very useful for the study of medium-to-long term (10-50 years) vegetation changes. To determine the quality of information derived from the classification process, accuracy assessment of the classification is implemented. Error matrix, which is primarily used in remote sensing for accuracy assessment, is typically based on an evaluation of the derived classification against some 'ground truth' or reference dataset. Regrettably 'ground truths' for some old historical photographs are rarely available. To solve this problem we formulate, in this Part I, methods of classification of aerial photographs and computation of accuracy assessment parameters of classification products in the absence of ground data, using the fuzzy classification technique. In addition, since point estimates of these accuracy parameters require associated standard errors, in order to be useful for statistical analysis, the method of computation of standard errors of accuracy measures using the bootstrap resampling techniques is presented. These methods are tested with historical aerial photographs, of part of Adulam Nature Reserve, Israel, spanning a period of 51 years. Results illustrate the applicability and efficiency of the proposed methods.

Original languageEnglish
Pages (from-to)153-176
Number of pages24
JournalInternational Journal of Remote Sensing
Volume27
Issue number1
DOIs
StatePublished - 10 Jan 2006

Fingerprint

Dive into the research topics of 'Methods for fuzzy classification and accuracy assessment of historical aerial photographs for vegetation change analyses. Part I: Algorithm development'. Together they form a unique fingerprint.

Cite this