Advanced threat assessment mechanisms are essential for the development of intelligent tactical decision-making systems. In response to the issues of insufficient real-time performance and low accuracy in traditional target threat assessment models, an enhanced version of the dynamic multi-swarm particle swarm optimization algorithm is proposed. This version is specifically tailored to refine the extreme learning machine (ELM) for evaluating threat levels of ground targets, designated as IDM-PSO-ELM. The innovation lies in integrating a dynamic multi-swarm strategy with the PSO framework, which is further augmented by incorporating Lévy flight, hybrid particles, and sine-cosine learning factor strategies. These enhancements significantly bolster the algorithm’s global exploration and local exploitation capabilities. The enhanced PSO algorithm is utilized to optimize the initial input weights and biases of the ELM model. The results show that the proposed model not only possesses higher accuracy in predicting target threats but also retains the high real-time performance characteristic compared with other assessment models.