基于SLSTM网络的两级修正机动目标跟踪方法

doi:10.19665/j.issn1001-2400.20241015

摘要/Abstract

摘要：

传统机动目标跟踪方法在机动模型建模方面,通过模型集自适应交互的方式,实现模型与目标真实运动的匹配。在跟踪非合作目标时,由于机动状态随时变化,且机动形式多样,当模型集内的有限个模型均无法精准表征其真实运动时,跟踪性能下降。将模型修正和状态修正两级神经网络融入到滤波递推过程中,提出一种基于堆叠长短时记忆(Stacked Long Short-Term Memory,SLSTM)网络的两级修正机动目标跟踪方法(Two Level Modified Maneuvering Target Tracking,TLM-MTT),第一级模型修正网络实时感知目标的机动,调整模型参数,实现机动模型的精准建模,第二级状态修正网络对状态估计进行实时补偿,提升滤波输出的精度。通过离线方式进行网络训练,训练后的网络用于在线实时跟踪,相较于传统方法和其他智能化滤波方法,文中所提方法对高机动目标跟踪具有更好的跟踪性能。

关键词: 目标跟踪, 长短时记忆网络, 卡尔曼滤波

Abstract:

In terms of maneuver model modeling,traditional maneuvering target tracking methods achieve matching between the model and the real motion of the target through adaptive interaction of the model set.When tracking non-cooperative targets,the maneuvering state changes at any time and the maneuvering forms are diverse.When the limited models in the model set cannot accurately represent its real motion,the tracking performance will degrade.This paper integrates the two level neural network of model correction and state correction into the filtering recursion process,and proposes a two-level modified maneuvering target tracking method(TLM-MTT) based on the stacked long short-term memory(SLSTM) network.The first-level model correction network perceives the maneuver of the target in real time,adjusts the model parameters,and realizes accurate modeling of the maneuver model.The second-level state correction network compensates the state estimation in real time to improve the accuracy of the filter output.The network is trained offline,and the trained network is used for online real-time tracking.Compared with traditional methods and other intelligent filtering methods,this method has a better tracking performance for high-maneuvering target tracking.

Key words: target tracking, long short-term memory(LSTM), kalman filter

中图分类号:

TN957.52

汪晋, 苏洪涛, 汪圣利, 陆超. 基于SLSTM网络的两级修正机动目标跟踪方法[J]. 西安电子科技大学学报, 2025, 52(1): 37-49.

WANG Jin, SU Hongtao, WANG Shengli, LU Chao. Two-level modified maneuvering target tracking method based on the SLSTM network[J]. Journal of Xidian University, 2025, 52(1): 37-49.

图/表 16

图1

图2

图3

图4

图5

表1

图6

表2

图7

图8

图9

图10

图11

表3

表4

表5

参考文献 21

[1]	LI X R, JILKOV V P. Survey of Maneuvering Target Tracking.Part Ⅰ.Dynamic Models[J]. IEEE Transactions on Aerospace and Electronic Systems, 2003, 39(4):1333-1364.
[2]	LI X R, JILKOV V P. Survey of Maneuvering Target Tracking.Part Ⅴ.Multiple-Model Methods[J]. IEEE Transactions on Aerospace and Electronic Systems, 2005, 41(4):1255-1321.
[3]	HU S, ZHAO W B, WANG J G. Interacting Multiple-Model Tracking Algorithm Fusing Input Estimation and Best Linear Unbiased Estimation Filter[J]. IET Radar,Sonar & Navigation, 2017, 11(1):70-77.
[4]	XU L F, LI X R, DUAN Z S. Hybrid Grid Multiple-Model Estimation with Application to Maneuvering Target Tracking[J]. IEEE Transactions on Aerospace and Electronic Systems, 2016, 52(1):122-136.
[5]	ANDREA W, ROLAND G, TOBIAS B. A Survey of Artificial Intelligence Approaches for Target Surveillance with Radar Sensors[J]. IEEE Aerospace and Electronic Systems Magazine, 2021, 36(7):26-43.
[6]	ALBERT H, MATIJS H, BASTIAAN D, et al. Deep Learning for Classification of Mini-UAVs Using Micro-Doppler Spectrograms in Cognitive Radar[J]. IEEE Aerospace and Electronic Systems Magazine, 2019, 34(11):46-56.
[7]	ERIK B, TIEN P, CHEE Y, et al. Machine Learning Artificial Intelligence for Sensor Data Fusion Opportunities and Challenges[J]. IEEE Aerospace and Electronic Systems Magazine, 2021, 36(7):80-93.
[8]	LINUS E, HENRY G, CHEN C, et al. Self-Supervised Representation Learning:Introduction,Advances and Challenges[J]. IEEE Signal Processing Magazine, 2022, 39(3):42-62.
[9]	贺王鹏, 胡德顺, 李诚, 等. 结合模板更新与轨迹预测的孪生网络跟踪算法[J]. 西安电子科技大学学报, 2024, 51(3):46-54.
	HE Wangpeng, HU Deshun, LI Cheng, et al. Siamese Network Tracking Using Template Updating and Trajectory Prediction[J]. Journal of Xidian University, 2024, 51(3):46-54.
[10]	宋建锋, 苗启广, 王崇晓, 等. 注意力机制的多尺度单目标跟踪算法[J]. 西安电子科技大学学报, 2021, 48(5):110-116.
	SONG Jianfeng, MIAO Qiguang, WANG Chongxiao, et al. Multi-Scale Single Object Tracking Based on the Attention Mechanism[J]. Journal of Xidian University, 2021, 48(5):110-116.
[11]	毛若昕, 张永权, 时贞云. 基于动态波门和DLSTM网络的机动多目标智能跟踪算法[J]. 现代雷达, 2023, 45(07):21-27.
	MAO Ruoxin, ZHANG Yongquan, SHI Zhenyun. Maneuvering Multi-Target Intelligent Tracking Algorithm Based on Dynamic Gate and DLSTM Network[J]. Modern Radar, 2023, 45(7):21-27.
[12]	KAZIMIERSKI W. Proposal of Neural Approach to Maritime Radar and Automatic Identification System Tracks Association[J]. IET Radar,Sonar & Navigation. 2016, 11(5):729-735.
[13]	LI S W, HU C, WANG R. A Maneuvering Tracking Method Based on LSTM and CS Model[J]. IEEE International Conference on Signal Information and Data Processing, 2019:1-4.
[14]	DENG L C, LI D, LI R F. Improved IMM Algorithm Based on RNNs[J]. Journal of Physics:Conference Series, 2020, 1518(1):012055.
[15]	张晓杰, 汪灏, 赵灵峰. 基于神经网络的多模型机动目标跟踪方法研究[J]. 无人系统技术, 2022, 5(2):71-79.
	ZHANG Xiaojie, WANG Hao, ZHAO Lingfeng. Research on Multi-Model Maneuvering Target Tracking Method Based on Neural Network[J]. Unmanned Systems Technology, 2022, 5(2):71-79.
[16]	MILAN A, REZATOFIGHI S H. Online Multi-Target Tracking Using Recurrent Neural Networks[C]// Proceedings of the Thirty-First AAAI Conference on Artificial Intelligence. Palo Alto: AAAI, 2017:4225-4232.
[17]	GAO C, LIU H W. Maneuvering Target Tracking with Recurrent Neural Networks for Radar Application[C]//2018 International Conference on Radar(RADAR). Piscataway:IEEE, 2018:1-8.
[18]	陈光武, 程鉴皓, 杨菊花, 等. 基于改进神经网络增强自适应UKF的组合导航系统[J]. 电子与信息学报, 2019, 41(7):1766-1773.
	CHEN Guangwu, CHENG Jianhao, YANG Juhua, et al. Improved Neural Network Enhanced Navigation System of Adaptive Unsented Kalman Filter[J]. Journal of Electronics & Information Technology, 2019, 41(7):1766-1773.
[19]	LIU J X, WANG Z L, XU M. DeepMTT:A Deep Learning Maneuvering Target-Tracking Algorithm Based on Bidirectional LSTM Network[J]. Information Fusion, 2020, 53:289-304.
[20]	熊伟, 朱洪峰, 崔亚奇. 在线学习的循环自适应机动目标跟踪算法[J]. 航空学报, 2022, 43(5):325250. doi: 10.7527/S1000-6893.2021.25250
	XIONG Wei,ZHU Hongfeng,CUI Yaqi. Recurrent Adaptive Maneuvering Target Tracking Algorithm Based on Online Learning[J]. Acta Aeronautica et Astronautica Sinica, 2022, 43(5):325250. doi: 10.7527/S1000-6893.2021.25250
[21]	何友, 修建娟, 刘瑜, 等. 雷达数据处理及应用[M]. 4版. 北京: 电子工业出版社, 2022.

场景	机动形式	最大过载/g	轨迹时长/s
1	蛇形	4.9	258
2	蛇形	9.0	210
3	爬升	3.5	330
4	爬升	5.0	160
5	俯冲	3.0	420
6	俯冲	5.9	350

场景	机动形式	最大过载/g	初始速度/(m·s^-1)	高度/m	备注
1	俯冲	6.0	220	4 000~6 000
2	爬升	5.5	200	500~1 000	低仰角模式
3	蛇形	2.0	190	3 000
4	圆周	1.5	80	300	低仰角模式

场景	IMM/m	RNN-MTT/m	DeepMTT/m	RKNN/m	TLM-MTT/m
1	22.3	17.4	16.8	17.1	12.1
2	15.0	12.2	11.6	11.8	11.0
3	7.0	6.4	6.8	6.8	6.2
4	10.6	8.2	8.0	8.2	7.8

场景	IMM/(°)	RNN-MTT/(°)	DeepMTT/(°)	RKNN/(°)	TLM-MTT/(°)
1	0.095	0.074	0.071	0.073	0.043
2	0.072	0.065	0.065	0.061	0.053
3	0.044	0.038	0.035	0.035	0.035
4	0.040	0.038	0.037	0.036	0.037

场景	IMM/(°)	RNN-MTT/(°)	DeepMTT/(°)	RKNN/(°)	TLM-MTT/(°)
1	0.092	0.066	0.043	0.044	0.031
2	0.185	0.157	0.155	0.152	0.140
3	0.043	0.040	0.040	0.038	0.037
4	0.129	0.094	0.087	0.090	0.082