Multilevel Image Segmentation Based on Fractional-Order Darwinian Particle Swarm Optimization

Hyperspectral remote sensing images contain hundreds of data channels. Due to the high dimensionality of the hyperspectral data, it is difficult to design accurate and efficient image segmentation algorithms for such imagery. In this paper, a new multilevel thresholding method is introduced for the segmentation of hyperspectral and multispectral images. The new method is based on fractional-order Darwinian particle swarm optimization (FODPSO) which exploits the many swarms of test solutions that may exist at any time. In addition, the concept of fractional derivative is used to control the convergence rate of particles. In this paper, the so-called Otsu problem is solved for each channel of the multispectral and hyperspectral data. Therefore, the problem of n-level thresholding is reduced to an optimization problem in order to search for the thresholds that maximize the between-class variance. Experimental results are favorable for the FODPSO when compared to other bioinspired methods for multilevel segmentation of multispectral and hyperspectral images. The FODPSO presents a statistically significant improvement in terms of both CPU time and fitness value, i.e., the approach is able to find the optimal set of thresholds with a larger between-class variance in less computational time than the other approaches. In addition, a new classification approach based on support vector machine (SVM) and FODPSO is introduced in this paper. Results confirm that the new segmentation method is able to improve upon results obtained with the standard SVM in terms of classification accuracies.