改进的K-Means算法在特征关联中的应用

摘要

Feature association is a critical step of radar emitter fusion identification through passive multi-sensor system.The feature association is based on the similarity of measurement data from the same sources of radiation.The measurement values of multiple radiation sources achieved by the passive multi-sen-sor system is classified and relevance determined and the feature information of the radiation sources is used to eliminate interconnected fuzzy.The selection of classification algorithms is an important part of the process of feature association.The K-Means clustering algorithm is based on the division algorithm and has been widely used in many fields.The proposed improved K-Means algorithm defines the similarity between sample points by grey relational degree instead of the traditional distance measure,then the simulation Radar-data-base sample set of radar emitter feature parameters is classified by the improved K-Means algorithm.Simula-tion results show that the improved K-Means algorithm improves the correct rate of association,but con-sumes more time.%特征关联是无源多传感器辐射源融合识别的一个关键步骤。特征关联是根据来源于同一辐射源的量测数据所具有的相似性，采用一定的算法和分配策略将多传感器获取的对多辐射源的量测值进行分类划分和关联判定，利用辐射源的特征信息来消除关联模糊。特征关联过程中一个重要环节就是分类算法的选取。K-Means算法是基于划分的聚类算法，已经广泛应用于诸多领域。改进了 K-Means算法，用灰关联度代替传统的距离度量定义了样本点间的距离，并对模拟产生的雷达辐射源特征参数样本集 Radar-da-tabase进行了分类。仿真结果表明，改进的K-Means算法提高了关联正确率，但消耗了更多时间。