【摘要】 针对天基红外预警分类识别问题，从典型火箭发动机喷焰光谱结构分析入手，以燃烧产物分子H₂O、 CO₂、 CO、 HCl特征波段辐射亮度比构造分类特征向量，提出以HCl和H₂O特征波段亮度比作为固、液类型大类的特征量，再以CO₂、 CO分别与H₂O特征波段亮度比为特征量进行固、液发动机子类进行分类的总体分类策略。采用粒子群优化的支持向量机分类算法对所提出的分类策略进行了数值实验验证，并探究了分类策略的抗噪性能。结果表明在该数值实验条件下，无噪声时平均分类正确率度达到100%,加入5%和10%的高斯噪声后，分类准确率在92%以上。更多还原

【Abstract】 To address the problem of classification and identification of space-based infrared warning, this paper starts from the analysis of the spectral structure of typical rocket exhaust plumes, and constructs classification feature vectors with the characteristic band radiation brightness ratios of combustion product molecules H₂O, CO₂, CO and HCl. The overall classification strategy of classifying solid and liquid engine subclasses by using HCl and H₂O characteristic band brightness ratios as the feature quantity of the solid and liquid type major class, and then using CO₂ and CO with H₂O characteristic band brightness ratios as the feature quantity, respectively, are proposed. The proposed classification strategy is verified by numerical experiments using a support vector machine classification algorithm with particle swarm optimization, and the noise immunity performance of the classification strategy is explored. The results show that under this numerical experimental condition, the average correct classification degree reaches 100% when there is no noise, and the classification accuracy is above 92% when 5% and 10% Gaussian noise is added.更多还原