基于量子克隆优化的SAR图像分类  

SAR Image Classification Based on Quantum Clonal Optimization

作　　者：马文萍[1] 焦李成[1] 张向荣[1] 李阳阳[1] 

机构地区：[1]西安电子科技大学智能信息处理研究所,陕西西安710071

出　　处：《电子学报》2007年第12期2241-2246,共6页Acta Electronica Sinica

基　　金：国家863项目(No.2006AA01Z107);国家973项目(No.2006CB705700);国家自然科学基金(No.672126)

摘　　要：将量子交叉操作引入人工免疫系统中的克隆选择优化,提出了一种用于解决SAR图像分类问题的量子克隆优化算法,基于Markov理论证明了其收敛性.新算法采用克隆选择操作同时在同一抗体周围的多个方向进行搜索,通过在各个子群体间采用量子交叉算子增强抗体间的信息交换,有效地克服了早熟现象.对X波段和Ku波段SAR图像的分类实验表明,与模糊C均值算法、K近邻算法和克隆选择算法相比,新算法的平均分类精度分别提高了13.57、11.79和5.79个百分点,而且新算法的鲁棒性也明显优于其他三种方法.Based on the clonal selection optimization with quantum crossover,a novel Quantum Clonal Optimization Algorithm is proposed for solving SAR image classification problems, theoretical analysis based on the theory of Markov has proved that the new algorithm could converage to the global optimum. The new algorithm can carry out searching in many directions around the same antibody simultaneously. The proposed quantum crossover operator realizes the information interactions among the sub-population so as to prevent premature convergence effectively. The experimental results on X-band and Ku-band SAR images indicate that compared with the Fuzzy C-means algorithm, the K-Nearest Neighbor algorithm, and the Clonal Selection Algorithm, the average conect rate of the new algorithm is improved by 13.57% ,11.79% and 5.79% ,and the robust of the new algorithm also outperforms the other three methods.

关 键 词：克隆选择 量子计算 SAR图像 图像分类 

分 类 号：TP18[自动化与计算机技术—控制理论与控制工程]