Fuzzy segmentation for geographic object-based image analysis

Abstract

Image segmentation partitions remote sensing images into image objects before assigning them to categorical land cover classes. Current segmentation methods require users to invest considerable time and effort in the search for meaningful image objects. As an alternative method we propose 'fuzzy' segmentation that offers more flexibility in dealing with remote sensing uncertainty. In the proposed method, original bands are processed using regression techniques to output fuzzy image regions which express degrees of membership to target land cover classes. Contextual properties of fuzzy regions can be measured to indicate potential spectral confusion. A 'defuzzification' process is subsequently conducted to produce the categorical land cover classes. This method was tested using data sets of both high and medium spatial resolution. The results indicate that this approach is able to produce classification with satisfying accuracy and requires very little user interaction.

Mapping and Characterizing Selected Canopy Tree Species at the Angkor World Heritage Site in Cambodia Using Aerial Data

“…Based on these heterogeneity criteria, the algorithm starts by building one-pixel objects. Adjacent individual pixel objects are merged with each other to form meaningful objects based on the heterogeneity criteria [ 27 ].…”