In recent years, drone swarms technology has demonstrated remarkable advantages in civil application fields such as, agricultural plant protection, and power inspection, thereby promoting the widespread adoption of multi-UAV collaborative operation modes. However, the high dynamicity and complexity inherent in UAV swarms have posed an urgent demand for real-time perception and precise management of their operational states, making real-time UAV swarm segmentation a key research focus. Mainstream clustering methods are limited by reliance on initialization and pre-set hyperparameters, unsuitable for unsupervised use in practical scenarios. This paper presents a novel unsupervised, Hyper-Parameter-Free (HPF) clustering algorithm based on multiple Gaussian functions’ summation properties, using peak count-standard deviation relations to find cluster centers. It also proposes a dual-criterion UAV swarm detection method using spatial positions and velocity directions. Real-dataset experiments show the method outperforms existing ones in detection accuracy and speed for swarm data, with performance similar to prior-knowledge-dependent algorithms.