基于支持向量回归与多核集成的红外成像导引头抗干扰性能评估方法  

Anti-interference Capability Evaluation Based on SVM Regression and Multi-kernel Boosting

作　　者：葛辰杰 陆志沣 洪泽华 马潮 余海鸣 赖鹏 乔宇 杨杰[1] GE Chenjie;LU Zhifeng;HONG Zehua;MA Chao;YU Haiming;LAI Peng;QIAO Yu;YANG Jie(School of Electronic Information and Electrical Engineering,Shanghai Jiao Tong University,Shanghai 200240,China;Shanghai Electro-Mechanical Engineering Institute,Shanghai 201109,China)

机构地区：[1]上海交通大学电子信息与电气工程学院,上海200240 [2]上海机电工程研究所,上海201109

出　　处：《上海航天》2019年第5期94-98,共5页Aerospace Shanghai

基　　金：航天先进技术联合研究中心技术创新项目基金(USCAST2015-13,USCAST2016-23,USCAST2016-8);上海航天科技创新基金(SAST2016008,SAST2017100,SAST2016085)

摘　　要：面对着日益复杂的对抗环境,红外成像导引头的抗干扰性能需要不断提高。如何全面、客观、准确地对红外成像导引头的抗干扰性能进行评估,是一项急需解决的难题。针对传统基于支持向量机的评估方法中单核学习能力的不足,提出了一种基于支持向量回归与多核集成的评估方法,该方法在抗干扰评估指标体系下得到了综合的抗干扰性能值,为红外成像导引头抗干扰性能评估提供了新的思路。该方法能够训练多个支持向量回归机并融合多个核函数的优势,充分利用了特征的多样性,进一步降低了回归误差。实验结果表明:该算法能够实现高效可靠的红外成像导引头抗干扰性能评估。Due to the increasingly complex fighting environment,there is a great need to improve the anti-interference capability.Therefore,how to evaluate the anti-interference capability of infrared guidance system precisely,objectively and efficiently has currently become an urgent issue.Since traditional SVM-based methods are unable to learn the regression effectively with single kernel machine,we propose an automatic method to evaluate the anti-interference capability based on SVM regression and multi-kernel boosting,using the manual index system and empirical evaluation results as training samples.The proposed method relies on multiple kernel learning to take full advantage of multiple features,which reduces the overall regression error.Experimental results show the effectiveness and efficiency of the proposed method on anti-interference capability evaluation.

关 键 词：红外成像导引头 性能评估 支持向量回归 多核学习 

分 类 号：V19[航空宇航科学与技术—人机与环境工程]