Segmentation of multispectral remote sensing images based on ant colony optimization algorithm

Abstract

Segmentation of remote sensing image is not only a hot topic but a difficult technological field in remote sensing image processing as well. Recently, Ant Colony Optimization (ACO) algorithm has been introduced into image segmentation. But seldom study has been done in segmentation of multispectral remote sensing images based on Ant Colony Optimization Algorithm. In this paper, ACO algorithm is used in segmentation of multispectral remote sensing images. Three vectors of multispectral remote sensing images at each pixel site are extracted as eigenvectors, such as multi-spectrum gray values at one pixel site, mean gray values of neighborhood pixels in each band, and multi-spectrum gradient values at one pixel site. They reflect both value features and spatial features of remote sensing images. The combination of these three eigenvectors is used as the fuzzy cluster features. Furthermore, ACO Algorithm is used to optimize fuzzy clustering process. This method not only improves the segmentation result of multispectral remote sensing images, but also controls calculation amount effectively. Experiment and comparison results show that fuzzy clustering algorithm optimized by ACO is a preferable mothod for segmentation of multispectral remote sensing images.