收稿日期:2021-08-08修回日期:2021-10-15基金项目:国家自然科学基金(62073124);国家航空基金资助项目(20170142002)作者简介:汪钰珠(1997-),男,河南驻马店人,硕士。研究方向:红外目标跟踪。*摘要:针对红外目标跟踪因遮挡产生的漂移问题,提出一种融合注意力机制和卡尔曼滤波的SiamRPN++红外目标跟踪算法,通过位置预测对遮挡目标进行跟踪。该算法引入注意力机制来增强SiamRPN++算法的置信度得分质量,使得遮挡判定更加稳定。为了实现目标被判遮挡后的有效跟踪,使用卡尔曼滤波对目标位置进行预测。实验结果表明,相比于原始算法SiamRPN++,该算法可以有效处理红外目标跟踪遮挡问题。关键词:目标跟踪;孪生网络;注意力机制;卡尔曼滤波;红外图像中图分类号:TP391.4文献标识码:ADOI:10.3969/j.issn.1002-0640.2022.11.005引用格式:汪钰珠,马建伟,吕进锋.基于位置预测的SiamRPN++红外目标跟踪算法[J].火力与指挥控制,2022,47(11):24-28.基于位置预测的SiamRPN++红外目标跟踪算法*汪钰珠,马建伟,吕进锋(河南科技大学,河南洛阳471000)InfraredTargetTrackingAlgorithminSiamRPN++BasedonPositionPredictionWANGYuzhu,MAJianwei,LYUJinfeng(HenanUniversityofScienceandTechnology,Luoyang471000,China)Abstract:Aimingattheproblemofinfraredtargettrackingdriftcausedbyocclusion,aSiamRPN++infraredtargettrackingalgorithmbasedonthefusionofattentionmechanismandKalmanfilterisproposed,whichcantracktheoccludedtargetthroughpositionprediction.Firstly,thealgorithmintroducestheattentionmechanismtoenhancetheconfidencescorequalityofSiamRPN++algorithm,makingtheocclusiondeterminationmorerobust.Then,inordertoachieveeffectivetrackingafterthetargetisjudgedtobeoccluded,Kalmanfilterisusedtopredictthetargetposition.TheexperimentalresultsshowthatcomparedwiththeoriginalalgorithmSiamRPN++,theproposedalgorithmcaneffectivelydealwiththeproblemofinfraredtargettrackingocclusion.Keywords:targettracking;siamesenetwork;attentionmechanism;Kalmanfilter;infraredimageCitationformat:WANGYZ,MAJW,LYUJF.Infraredtargettrackingalgorithminsiamrpn++basedonpositionprediction[J].FireControl&CommandControl,2022,47(11):24-28.0引言随着红外摄像头在成...