A GA-based sequential fuzzy segmentation approach for classification of remote sensing images

In this paper we suggest an integrated spectral-spatial classification scheme for handling remotely sensed images. The method combines the results of supervised pixel-based classification with spatial information from unsupervised image segmentation. Pixel wise classification is implemented here by a fuzzy rule-based classifier using spectral/textural features of pixels. In regard to the image segmentation task, the following novelties are introduced. First, we apply a robust kernelized Fuzzy c-means clustering algorithm with spatial constraints (KFCM_S) to generate a fuzzy membership map. Operating on this transformed space, a Genetic Sequential Image Segmentation (GeneSIS) algorithm is next developed to partition the image into homogeneous regions. GeneSIS follows a sequential object extraction approach whereby at each iteration, a single object is extracted by invoking a GA-based object extraction algorithm. This module evaluates the fuzzy content of candidate regions, and through an effective fitness function design provides objects with optimal balance between fuzzy region coverage and consistency. The final classification results are obtained by performing fuzzy majority voting on the fuzzy degrees derived from pixel classifier over the segments obtained by GeneSIS. The validity of the proposed method is shown on the land cover classification of an agricultural area using a high-resolution IKONOS image, in terms of image segmentation quality and accuracy improvements compared to pixel wise classifiers.