A Hybrid PSO-ISODATA Algorithm for Remote Sensing Image Segmentation

Remote sensing image segmentation is a very important process in the remote sensing information extraction area. In order to avoid disadvantages of the current algorithms, this paper proposed a new algorithm combined PSO and ISODATA. The hybrid PSO-ISODATA algorithm first changes the color space of the images. Then, the initial cluster number is determined by the combined algorithm, and finally, the automatic segmentation of remote sensing images is achieved through multiple iterations. Many segmentation experiments on different spatial resolution remote sensing images using the proposed methods in this paper, we also compared the method to the current existing methods such as K-means, PSO, ISODATA, and PSO-K-means methods. The results show that the hybrid PSO-ISODATA algorithm can determine the initial cluster number adaptively, avoid the local optima of K-means and ISODATA algorithms, increase the searching capability of PSO, and the segmentation results are much more close to the actual situation.