基于连通检测和区域扩展的多重聚类方法研究

针对现有聚类算法在分选过程中存在辐射源定位和脉冲划分的困难，提出了基于连通检测和区域扩展的多重聚类方法。使用网格统计的方法对脉冲列分布密度进行量化，再借鉴图像分割算法，将量化的网格矩阵划分连通区域。基于峰值检测和梯度下降的原则扩展连通区域，通过边缘检测得到最终聚类结果。仿真结果表明，该算法能保证更多的聚类脉冲列参与参数分选，提高辐射源识别率。

Multiple clustering method based on connectivity detection and regional expansion

For the difficult of source positioning and pulse partition of existing clustering algorithms in the process of sorting , the clustering method based on the detection and regional expansion is put forward . Statistical methods use the grid to quantify the distribution density of the pulse train . By using image segmentation algorithm ,quantization matrix grid is divided into connectivity region .The connectivity region is extended based on a peak detection and gradient descent principle ,and the final result is given by the edge de‐tection .Simulation results show that the clustering algorithm can ensure that more pulse train parameters involved in sorting ,and inprove recognition rate .