PARAMETER TESTS FOR IMAGE SEGMENTATION OF AN AGRICULTURAL REGION

Abstract

Segmentation, mainly the operation of object extraction from the total image, is the initial and a very important step of object-based image processing since it directly affects the performance of the processing, quality of the product and the accuracy of the results. In this study, the segmentation parameters of Definiens, widely used GEOBIA software, were tested by comparing the segments obtained by various combinations of values. As the study area, an agricultural region was selected and the results were evaluated in scope of extracting the field boundaries. The tests were conducted by producing 27 segmented images of the SPOT 4 data acquired on April 26th, 2007. The performances for the given conditions (combination of criteria) were compared to each other and the criteria of value selection that can be used for different aims were outlined. The results can be used in further studies for estimating the optimum values in accordance with the purpose.