GMC: Generating Confusion Matrix from data ArcGIS®
ISBN 978-85-88783-11-9
Authors
1Prina, B.Z.; 2Trentin, R.
1UNIVERSIDADE FEDERAL DE SANTA MARIA Email: brunozprina@gmail.com
2UNIVERSIDADE FEDERAL DE SANTA MARIA Email: romario.trentin@gmail.com
Abstract
The aim of this paper is to create a methodology for analyzing the result a digital image classification using the app ArcGIS®, as data source. With this proposal, are used two apps developed with Visual Basic programming language: the GMC (which will generate the confusion matrix) and AVACIM (to calculate the results of the confusion matrix). As a “test area” to research, was mapped a random polygon in an image that included varied uses of soil (water, vegetation, bare soil, farming and countryside). Like this, we choose one part of the scene of an image sensor Landsat-8, the city of Fortaleza dos Valos/Brazil. The classification was performed in ArcGIS®, by collecting samples of different uses of soil. Later, was applied the “Intersect” algorithm (from ArcGIS®) generating a new file, with the correlation existing between points classified (your classes) with these same points located in the final image (classified). After, was organized in a database (of Access®) to continue with the methodological procedures in the GMC app. With the GMC, was generated a confusion matrix, which was saved, automatically, in another database (Access®). The last stage occurred in the AVACIM app, calculating a confusion matrix, to get the results about Kappa index, with the aim of to assess the quality of image classification. Thus, in the final classification was obtained a Kappa index of 0.9513, rated “Excellent”, evidencing, thus that samples collected were consistent with classified information. By and large, considered very important the method of evaluation a digital image classification exposed in this paper, by the fact that, through the development of two apps was obtained quantify and evaluate the result of the classification using the ArcGIS® app, as a source of primary information.
Keywords
Digital Image Processing; Use of soil; Visual Basic