基于聚类和随机参考分类器的对海雷达目标识别算法

Target Recognition Algorithm for Maritime Surveillance Radars Based on Clustering and Random Reference Classifier

为了进一步提高复杂干扰环境下对海雷达目标识别的泛化能力,提出基于k-medoids聚类和随机参考分类器(RRC)的动态选择集成算法(KMRRC).主要利用重采样技术生成多个基分类器,然后基于成对多样性度量准则将基分类器划分为多个簇,并基于校验数据集为每个基分类器构建相应的RRC模型,最后利用RRC从各个簇中动态选择竞争力最强的部分基分类器进行集成决策.通过寻优实验确定KMRRC的参数设置,随后利用Java调用Weka API在自建的目标全极化高分辨距离像(HRRP)样本库及17个UCI数据集上进行KMRRC与常用的9种集成算法和基分类算法的对比实验,并进一步研究多样性度量方法的选取对KMRRC性能的影响.实验验证文中算法在对海雷达目标识别领域的有效性.

To improve the generalization ability of maritime surveillance radars in complicatedly interferential environment, a dynamic ensemble selection algorithm based on k-medoids clustering and random reference classifier（KMRRC） is proposed. Firstly, a pool of base classifiers are generated through Bagging technique. Secondly, k-medoids clustering is used to divided all the base classifiers into several clusters based on pairwise diversity metric. Then, the RRC model for each base classifier is constructed on the basis of validation dataset. Finally, the RRC model is employed to select some of the most competent classifiers from each cluster for ensemble and decision making. The parameters of KMRRC are determined by optimization experiment based on the self-buih high resolution range profile（HRRP） dataset, and the performance of KMRRC is compared with nine ensemble methods and the base classification algorithm using the HRRP dataset and other seventeen UCI datasets in Java environment with a Weka stand-alonelibrary. Besides, the influence of the diversity measures on the performance of KMRRC is further studied. The feasibility of KMMRRC in the field of target recognition for maritime surveillance radars is verified by experiments.