Target Cluster Partitioning of Density Peak Clustering Algorithm Based on Boundary Detection

Aiming at the problem that the traditional Density Peak Clustering (DPC) algorithm has insufficient ability to identify boundary points when applied to clustering and partitioning of imbalanced cluster targets in complex air combat scenarios, this paper proposes a boundary detection clustering algorithm that integrates asymmetric metrics and local density. It introduces k-nearest neighbor sampling to evaluate the spatial distribution characteristics of data points, and constructs a decision graph combined with the minimum distance to achieve accurate identification and clustering of boundary points of imbalanced clusters. The goal is to quickly detect the splitting/merging of clusters during the movement of air cluster targets, determine the cluster partitioning, and obtain preliminary identification results of the clusters. The clustering results of splitting/merging can be applied to fields such as air battlefield cluster type identification, threat assessment and prediction, and cluster target tracking algorithm research, effectively improving the accuracy and robustness of cluster clustering.