The Modified Clonal Selection Algorithm Applied to the Remote Sensing Image Information Extracting

Clonal selection algorithm is one of the important artificial intelligence algorithms, which used as a powerful information processing and problem solving paradigm in both the scientific and engineering fields. However, few studies concern application of CLONALG method in information extracting of satellite image. In this paper, we suggest modified Clonal selection algorithm (M-CLONALG) improved by niche technology´s sharing function and memory calculator technique applying to remote sensing (RS) image information extracting. In order to testify the correctness that M-CLONALG improves the stability of searching results, accuracy of the global optimization, we compare with maximum likelihood method, minimum distance method and normal CLONALG method. Experimental results confirm that M-CLONALG has self-organizing, self-learning ability, and no limitation to the distribution of training samples from the global data, with complete convergence, can quickly search the best center of classification clustering at high accuracy. Therefore, our method M-CLONALG is superior to other three algorithms in remote sensing image information extracting, and its overall accuracy and Kappa statistic reach 91.7% and 0.89 respectively.