基于位置预测的SiamRPN++红外目标跟踪算法  

Infrared Target Tracking Algorithm in SiamRPN++ Based on Position Prediction

作　　者：汪钰珠 马建伟[1] 吕进锋 WANG Yuzhu;MA Jianwei;LYU Jinfeng(Henan University of Science and Technology,Luoyang 471000,China)

机构地区：[1]河南科技大学,河南洛阳471000

出　　处：《火力与指挥控制》2022年第11期24-28,共5页Fire Control & Command Control

基　　金：国家自然科学基金(62073124);国家航空基金资助项目(20170142002)。

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。Aiming at the problem of infrared target tracking drift caused by occlusion,a SiamRPN++ infrared target tracking algorithm based on the fusion of attention mechanism and Kalman filter is proposed,which can track the occluded target through position prediction. Firstly,the algorithm introduces the attention mechanism to enhance the confidence score quality of SiamRPN++ algorithm,making the occlusion determination more robust. Then,in order to achieve effective tracking after the target is judged to be occluded,Kalman filter is used to predict the target position. The experimental results show that compared with the original algorithm SiamRPN ++,the proposed algorithm can effectively deal with the problem of infrared target tracking occlusion.

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 

分 类 号：TP391.4[自动化与计算机技术—计算机应用技术]