针对信息融合中目标威胁估计的特点，分析了传统目标威胁估计方法和支持向量机(SVM)的不足。采用粒子群算法(PSO)对SVM中惩罚参数c和核函数g进行优化，建立了改进的SVM(PSO_SVM)目标威胁估计模型及算法。介绍了粒子群算法和支持向量机的原理，建立了一种新的PSO_SVM目标威胁估计模型；基于该模型，实现了PSO_SVM目标威胁估计算法。为适应该算法，对数据进行了预处理，包括数据量化和归一化。交叉验证寻找最佳参数时，采用PSO算法进行优化。采集75组原始数据用于仿真实验，其中60组作为训练集，15组作为测试集。仿真实验表明，该算法预测误差为0，达到了预期目标。实验结果真实、准确地反映了实际情况，证明了该方法的有效性。

On the basis of the characteristics of target threat assessment in information fusion, the weaknesses of traditional methods for target threat assessment and Support Vector Machine (SVM) were analyzed. By using the Particle Swarm Optimization (PSO) to optimize the penalty parameter c and core function g in the SVM, a new target threat assessment model (PSO_SVM) was established and the PSO_SVM algorithm was achieved based on the model. To satisfy the requirements of PSO_SVM algorithm, data was preprocessed, including quantification and normalization. When crossvalidation method was used to find the best parameters, the POD was used for network training. 75 group data were used in simulation experiments, among them 60 group data were train sets and the others were test sets. Experimental results show that the error of the PSO_SVM method is 0, reaching the desired goal, which proves the accuracy and efficiency of the proposed method.