基于自然最近邻和霍夫变换的密度峰值聚类多目标定位算法

Multi-target Localization Algorithm based on Density Peak Clustering using Natural Nearest Neighbor and Hough Transform

摘要:

针对多站多目标AOA无源定位中目标检测率低与虚假交点干扰问题，开展了基于密度峰值聚类的定位算法研究；通过引入自然最近邻（NNN）方法自适应确定数据点邻域范围，结合截断核密度与高斯核密度改进局部密度计算模型，有效提升了算法对非均匀密度数据的适应性；在此基础上，采用霍夫变换（HT）自动检测决策值分布中的直线特征，实现了聚类中心的全自动提取；仿真实验表明，在测向误差范围内，定位精度提升10%-30%，复杂环境目标检测率提高10%-20%，高误差场景下定位误差仅为基准线法的65%，关联正确率达92%；实验结果验证了该算法在非理想环境中的鲁棒性优势，通过优化邻域选择机制和密度计算模型，有效减少了人工参数干预，为多目标无源定位提供了一种高效可靠的解决方案。

Abstract:

Aiming at the problem of low target detection rate and false intersection interference in multi station multi-target AOA passive positioning, a positioning algorithm based on density peak clustering was studied; By introducing the natural nearest neighbor (NNN) method to adaptively determine the neighborhood range of data points, and combining truncated kernel density and Gaussian kernel density to improve the local density calculation model, the adaptability of the algorithm to non-uniform density data is effectively enhanced; On this basis, the Hough transform (HT) is used to automatically detect the linear features in the decision value distribution, achieving fully automatic extraction of clustering centers; Simulation experiments show that within the range of direction finding error , the positioning accuracy is improved by 10% -30%, and the target detection rate in complex environments is improved by 10% -20%. In high error scenarios, the positioning error is only 65% of the baseline method, and the correlation accuracy reaches 92%; The experimental results have verified the robustness advantage of the algorithm in non ideal environments. By optimizing the neighborhood selection mechanism and density calculation model, it effectively reduces manual parameter intervention and provides an efficient and reliable solution for multi-target passive localization.