基于Transformer复杂运动辨识的机动星凸形扩展目标跟踪方法

陈辉 边斌超 连峰 韩崇昭

陈辉, 边斌超, 连峰, 等. 基于Transformer复杂运动辨识的机动星凸形扩展目标跟踪方法[J]. 雷达学报(中英文), 待出版. doi: 10.12000/JR24031
引用本文: 陈辉, 边斌超, 连峰, 等. 基于Transformer复杂运动辨识的机动星凸形扩展目标跟踪方法[J]. 雷达学报(中英文), 待出版. doi: 10.12000/JR24031
CHEN Hui, BIAN Binchao, LIAN Feng, et al. A novel method for tracking complex maneuvering star convex extended targets using transformer network[J]. Journal of Radars, in press. doi: 10.12000/JR24031
Citation: CHEN Hui, BIAN Binchao, LIAN Feng, et al. A novel method for tracking complex maneuvering star convex extended targets using transformer network[J]. Journal of Radars, in press. doi: 10.12000/JR24031

基于Transformer复杂运动辨识的机动星凸形扩展目标跟踪方法

doi: 10.12000/JR24031
基金项目: 国家自然科学基金(62163023, 61873116, 62173266, 62366031),甘肃省教育厅产业支撑计划项目(2021CYZC-02),2023年甘肃省军民融合发展专项资金项目,2024年甘肃省重点人才项目资助
详细信息
    作者简介:

    陈 辉,教授,博士生导师,主要研究方向为数据融合、统计信号处理、机器学习和智能决策

    边斌超,硕士生,主要研究方向为神经网络和扩展目标跟踪

    连 峰,教授,博士生导师,主要研究方向为多源信息融合、滤波与估计算法、气动融合算法

    韩崇昭,教授,博士生导师,主要研究方向为数据融合、电子对抗、雷达目标跟踪等

    通讯作者:

    陈辉 chenh@lut.edu.cn

    边斌超 bianbinchao@lut.edu.cn

  • 责任主编:李天成 Corresponding Editor: LI Tiancheng
  • 中图分类号: TP389.1

A Novel Method for Tracking Complex Maneuvering Star Convex Extended Targets Using Transformer Network

Funds: The National Natural Science Foundation of China (62163023, 61873116, 62173266, 62366031), The Industrial Support Project of Education Department of Gansu Province (2021CYZC-02), The Special Funds Project for Civil-Military Integration Development of Gansu Province in 2023, The Key Talent Project of Gansu Province in 2024
More Information
  • 摘要: 针对复杂的机动扩展目标跟踪问题,利用Transformer网络设计了一种有效的星凸不规则形状机动扩展目标跟踪方法。首先,该文研究利用alpha-shape算法建立了星凸形状的变化模型,实现了静态场景下的星凸形扩展目标的形状估计。然后,通过对目标状态转移矩阵进行重新设计,结合Transformer网络对机动扩展目标运动状态转移矩阵进行实时估计,实现了对复杂机动目标运动过程的精准跟踪。进一步地,将估计得到的形状轮廓与运动状态进行融合,最终实现了对星凸形机动扩展目标的实时跟踪。最后,通过构造复杂的机动扩展目标跟踪场景,利用多重性能指标测试算法对形状和运动状态的综合估计性能,验证了算法的有效性。

     

  • 图  1  产生的目标量测点

    Figure  1.  Generated target measurements

    图  2  利用alpha-shape算法提取目标形状

    Figure  2.  Target shape extraction using alpha-shape algorithm

    图  3  静态场景下的轮廓估计

    Figure  3.  Contour estimation in static scenes

    图  4  神经网络模型整体结构

    Figure  4.  Overall structure of the neural network model

    图  5  一维特征信息提取

    Figure  5.  1-D feature information extraction

    图  6  特征信息提取结构

    Figure  6.  Feature information extraction structure

    图  7  全连接层结构

    Figure  7.  Full connect layer

    图  8  训练过程中损失值变化情况

    Figure  8.  Changes in loss values during training

    图  9  网络模型训练所用的数据集

    Figure  9.  Dataset used for network model training

    图  10  星凸形机动扩展目标跟踪方法

    Figure  10.  Star convex maneuvering extended target tracking

    图  11  算法可行性测试

    Figure  11.  Algorithm feasibility testing

    图  12  可行性测试的面积误差对比

    Figure  12.  Comparison of area errors in feasibility tests

    图  13  可行性测试的IOU对比

    Figure  13.  Comparison of IOU in feasibility tests

    图  14  可行性测试中的FDA对比

    Figure  14.  Comparison of FDA in feasibility tests

    图  15  形状自适应性测试结果

    Figure  15.  Shape adaptability test results

    图  16  形状自适应性测试的面积误差对比

    Figure  16.  Comparison of area errors for the shape adaptation tests

    图  17  形状自适应性测试的IOU对比

    Figure  17.  IOU comparison for shape adaptation tests

    图  18  形状自适应性测试的FDA对比

    Figure  18.  FDA comparison for shape adaptation tests

    图  19  不同算法的单步运行时间对比

    Figure  19.  Comparison of single-step running time for different algorithms

    图  20  鲁棒性测试

    Figure  20.  Robustness test

    1  结合Transformer的星凸形机动扩展目标跟踪部分伪代码

    1.   The pseudo-code of star convex maneuvering extended target tracking using Transformer algorithm

     输入:$ {A_1},{A_2},{x_0},{P_0},{W_k},{U_k},{Q_k},{R_k},P_0^{{\text{cap}}} $
     步骤1:预测
      for k = 1: steps
       % 量测集处理与形状初步处理
        ${{\boldsymbol{\bar z}}_k} = \dfrac{1}{l}\displaystyle\sum\limits_{i = 1}^l {{{\boldsymbol{z}}_{k,l}}} $
        ${{\boldsymbol{\tilde Z}}_k} = \{ {{\boldsymbol{\tilde z}}_{k,l}}\} $
        ${{\boldsymbol{C}}_k} = {{\mathrm{as}}} (a,{{\boldsymbol{\tilde Z}}_k})$
        ${{\boldsymbol{Z}}^k} = {{\mathrm{s}}} ({{\boldsymbol{C}}_k})$
       % 静态形状预测
        $ \mathcal{X}_k^ - - {\bar{\mathcal{X}} _k} = {\boldsymbol{A}}(\mathcal{X}_{k - 1}^ - - {\bar {\mathcal{X}} _k}) $
        ${\boldsymbol{P}}_k^ - = {\boldsymbol{A}}{{\boldsymbol{P}}_k}{{\boldsymbol{A}}^{\text{T}}} + {{\boldsymbol{W}}_k}{\boldsymbol{W}}_k^{\text{T}}$
       % 运动状态预测
        $\hat \chi _{k - 1}^j = {\boldsymbol{F}}_k^i\chi _{k - 1}^j$
        $ {{\boldsymbol{x}}_{k|k - 1}} = \dfrac{1}{m}\displaystyle\sum\limits_{j = 1}^m {\chi _{k - 1}^j} $
       % 形态预测
        $\mathcal{X}_{k,s}^ - = \mathcal{X}_k^ - + {{\boldsymbol{x}}_{k|k - 1}}$
       end
     步骤2:更新
       for k = 1: steps
       % 静态形状更新
        ${{\boldsymbol{K}}_k} = {\boldsymbol{P}}_k^ - {{\boldsymbol{H}}^{\text{T}}}{({{\boldsymbol{S}}_k}{\boldsymbol{HP}}_k^ - {{\boldsymbol{H}}^{\text{T}}} + {\boldsymbol{I}})^{ - 1}}$
        ${\hat {\mathcal{X}}_k} = \mathcal{X}_k^ - + {{\boldsymbol{K}}_k}{{\boldsymbol{Z}}^k}$
        ${{\boldsymbol{P}}_k} = ({\boldsymbol{I}} - {{\boldsymbol{K}}_k}{{\boldsymbol{S}}_k}{\boldsymbol{H}}){\boldsymbol{P}}_k^ - $
       % 状态转移矩阵更新
        ${{\boldsymbol{F}}_k} = {\text{TFMETT}}({{\boldsymbol{\bar z}}_k})$
       % 运动状态更新
        ${{\boldsymbol{\hat x}}_k} = {{\boldsymbol{x}}_{k|k - 1}} + {\boldsymbol{K}}_k^{{\text{cap}}}({{\boldsymbol{\bar z}}_k} - {{\boldsymbol{z}}_{k|k - 1}})$
        ${\boldsymbol{P}}_k^{{\text{cap}}} = {\boldsymbol{P}}_{k|k - 1}^{{\text{cap}}} - {\boldsymbol{K}}_k^{{\text{cap}}}{\boldsymbol{S}}_k^{{\text{cap}}}{\left( {{\boldsymbol{K}}_k^{{\text{cap}}}} \right)^{\text{T}}}$
       % 形态更新
        ${\hat {\mathcal{X}}_{k,s}} = {\hat {\mathcal{X}}_k} + {{\boldsymbol{\hat x}}_k}$
       end
     输出:${\hat {\mathcal{X}}_{k,s}}$
    下载: 导出CSV

    表  1  50次蒙特卡罗仿真测试中各种算法的平均单步运行时间

    Table  1.   Average single-step running time of various algorithms in 50 Monte Carlo simulation tests

    算法平均单步运行时间(s)
    RHM0.38
    GPR0.41
    TFMETT0.26
    下载: 导出CSV
  • [1] GRANSTRÖM K and BAUM M. A tutorial on multiple extended object tracking[EB/OL]. https://www.techrxiv.org/doi/full/10.36227/techrxiv.19115858.v1, 2022.
    [2] LAN Jian. Extended object tracking using random matrix with extension-dependent measurement numbers[J]. IEEE Transactions on Aerospace and Electronic Systems, 2023, 59(04): 4464–4477. doi: 10.1109/TAES.2023.3241888.
    [3] MANNARI P, THARMARASA R, and KIRUBARAJAN T. Extended target tracking under multitarget tracking framework for convex polytope shapes[J]. Signal Processing, 2024, 217: 109321. doi: 10.1016/j.sigpro.2023.109321.
    [4] TAN Jintao, QI Guoqing, QI Junjie, et al. Model parameter adaptive approach of extended object tracking using random matrix and identification[C]. 2022 International Conference on Cyber-Physical Social Intelligence, Nanjing, China, 2022: 91–97. doi: 10.1109/ICCSI55536.2022.9970662.
    [5] BAUR T, REUTER J, ZEA A, et al. Extent estimation of sailing boats applying elliptic cones to 3D LiDAR data[C]. 2022 25th International Conference on Information Fusion (FUSION), Linksping, Sweden, 2022: 1–8. doi: 10.23919/FUSION49751.2022.9841265.
    [6] ZHANG Yongquan, JI Hongbing, and HU Qi. A box-particle implementation of standard PHD filter for extended target tracking[J]. Information Fusion, 2017, 34: 55–69. doi: 10.1016/j.inffus.2016.06.007.
    [7] ZHANG Xiaoxiao and LAN Jian. Measurement combination estimator for multisensor extended object tracking using random matrix[J]. IEEE Transactions on Aerospace and Electronic Systems, 2024, 60(1): 698–715. doi: 10.1109/TAES.2023.3329075.
    [8] BAUM M and HANEBECK U D. Random hypersurface models for extended object tracking[C]. 2009 IEEE International Symposium on Signal Processing and Information Technology (ISSPIT), Ajman, UAE, 2009: 178–183. doi: 10.1109/ISSPIT.2009.5407526.
    [9] BAUM M and HANEBECK U D. Shape tracking of extended objects and group targets with star-convex RHMs[C]. 14th International Conference on Information Fusion, Chicago, USA, 2011: 1–8.
    [10] YANG Jinlong, LI Peng, and GE Hongwei. Extended target shape estimation by fitting B-spline curve[J]. Journal of Applied Mathematics, 2014, 2014: 741892. doi: 10.1155/2014/741892.
    [11] KOCH J W. Bayesian approach to extended object and cluster tracking using random matrices[J]. IEEE Transactions on Aerospace and Electronic Systems, 2008, 44(3): 1042–1059. doi: 10.1109/TAES.2008.4655362.
    [12] FELDMANN M, FRÄNKEN D, and KOCH W. Tracking of extended objects and group targets using random matrices[J]. IEEE Transactions on Signal Processing, 2011, 59(4): 1409–1420. doi: 10.1109/TSP.2010.2101064.
    [13] SWAIN A and CLARK D. Extended object filtering using spatial independent cluster processes[C]. 2010 13th International Conference on Information Fusion, Edinburgh, UK, 2010: 1–8. doi: 10.1109/ICIF.2010.5711886.
    [14] ABDALLAH F, GNING A, and BONNIFAIT P. Box particle filtering for nonlinear state estimation using interval analysis[J]. Automatica, 2008, 44(3): 807–815. doi: 10.1016/j.automatica.2007.07.024.
    [15] LAN Jian and LI Xiaorong. Tracking of maneuvering non-ellipsoidal extended object or target group using random matrix[J]. IEEE Transactions on Signal Processing, 2014, 62(9): 2450–2463. doi: 10.1109/TSP.2014.2309561.
    [16] LAN Jian and LI Xiaorong. Tracking of extended object or target group using random matrix—part I: New model and approach[C]. 2012 15th International Conference on Information Fusion, Singapore, Singapore, 2012: 2177–2184.
    [17] WAHLSTROM N and ÖZKAN E. Extended target tracking using Gaussian processes[J]. IEEE Transactions on Signal Processing, 2015, 63(16): 4165–4178. doi: 10.1109/TSP.2015.2424194.
    [18] LI Xiaorong and ZHANG Youmin. Multiple-model estimation with variable structure. V. Likely-model set algorithm[J]. IEEE Transactions on Aerospace and Electronic Systems, 2000, 36(2): 448–466. doi: 10.1109/7.845222.
    [19] LI Xiaorong, ZWI X, and ZWANG Y. Multiple-model estimation with variable structure. III. Model-group switching algorithm[J]. IEEE Transactions on Aerospace and Electronic Systems, 1999, 35(1): 225–241. doi: 10.1109/7.745694.
    [20] JILKOV V P, ANGELOVA D S, and SEMERDJIEV T A. Design and comparison of mode-set adaptive IMM algorithms for maneuvering target tracking[J]. IEEE Transactions on Aerospace and Electronic Systems, 1999, 35(1): 343–350. doi: 10.1109/7.745704.
    [21] 王昱淇, 卢宙, 蔡云泽. 基于一致性的分布式变结构多模型方法[J]. 自动化学报, 2021, 47(7): 1548–1557. doi: 10.16383/j.aas.c190091.

    WANG Yuqi, LU Zhou, and CAI Yunze. Consensus-based distributed variable structure multiple model[J]. Acta Automatica Sinica, 2021, 47(7): 1548–1557. doi: 10.16383/j.aas.c190091.
    [22] 赵楚楚, 王子微, 丁冠华, 等. 基于模糊逻辑的改进自适应IMM跟踪算法[J]. 信号处理, 2021, 37(5): 724–734. doi: 10.16798/j.issn.1003-0530.2021.05.005.

    ZHAO Chuchu, WANG Ziwei, DING Guanhua, et al. Fuzzy-logic adaptive IMM algorithm for target tracking[J]. Journal of Signal Processing, 2021, 37(5): 724–734. doi: 10.16798/j.issn.1003-0530.2021.05.005.
    [23] LIU Jingxian, YANG Shuhong, and YANG Fan. A cross-and-dot-product neural network based filtering for maneuvering-target tracking[J]. Neural Computing and Applications, 2022, 34(17): 14929–14944. doi: 10.1007/s00521-022-07338-7.
    [24] SONG Fei, LI Yong, CHENG Wei, et al. An improved Kalman filter based on long short-memory recurrent neural network for nonlinear radar target tracking[J]. Wireless Communications and Mobile Computing, 2022, 2022: 8280428. doi: 10.1155/2022/8280428.
    [25] GAO Chang, LIU Hongwei, ZHOU Shenghua, et al. Maneuvering target tracking with recurrent neural networks for radar application[C]. 2018 International Conference on Radar (RADAR), Brisbane, Australia, 2018: 1–5. doi: 10.1109/RADAR.2018.8557284.
    [26] YU Wanting, YU Hongyi, DU Jianping, et al. DeepGTT: A general trajectory tracking deep learning algorithm based on dynamic law learning[J]. IET Radar, Sonar & Navigation, 2021, 15(9): 1125–1150. doi: 10.1049/rsn2.12111.
    [27] LIU Jingxian, WANG Zulin, and XU Mai. DeepMTT: A deep learning maneuvering target-tracking algorithm based on bidirectional LSTM network[J]. Information Fusion, 2020, 53: 289–304. doi: 10.1016/j.inffus.2019.06.012.
    [28] JOUABER S, BONNABEL S, VELASCO-FORERO S, et al. NNAKF: A neural network adapted Kalman filter for target tracking[C]. ICASSP 2021–2021 IEEE International Conference on Acoustics, Speech and Signal Processing (ICASSP), Toronto, Canada, 2021: 4075–4079. doi: 10.1109/ICASSP39728.2021.9414681.
    [29] AFTAB W and MIHAYLOVA L. A learning Gaussian process approach for maneuvering target tracking and smoothing[J]. IEEE Transactions on Aerospace and Electronic Systems, 2021, 57(1): 278–292. doi: 10.1109/TAES.2020.3021220.
    [30] NEZHADARYA E, LIU Yang, and LIU Bingbing. BoxNet: A deep learning method for 2d bounding box estimation from bird's-eye view point cloud[C]. 2019 IEEE Intelligent Vehicles Symposium (IV), Paris, France, 2019: 1557–1564. doi: 10.1109/IVS.2019.8814058.
    [31] STEUERNAGEL S, THORMANN K, and BAUM M. CNN-based shape estimation for extended object tracking using point cloud measurements[C]. 2022 25th International Conference on Information Fusion (FUSION), Linksping, Sweden, 2022: 1–8. doi: 10.23919/FUSION49751.2022.9841254.
    [32] BARINGOLTS T V, DOMIN D V, ZHUK S Y, et al. Adaptive algorithm of maneuvering target tracking in complex jamming situation for multifunctional radar with phased antenna array[J]. Radioelectronics and Communications Systems, 2019, 62(7): 342–355. doi: 10.3103/S0735272719070021.
    [33] XU Xiaolong and HARADA K. Automatic surface reconstruction with alpha-shape method[J]. The Visual Computer, 2003, 19(7): 431–443. doi: 10.1007/s00371-003-0207-1.
    [34] EDELSBRUNNER H, KIRKPATRICK D, and SEIDEL R. On the shape of a set of points in the plane[J]. IEEE Transactions on Information Theory, 1983, 29(4): 551–559. doi: 10.1109/TIT.1983.1056714.
    [35] 李世林, 李红军. 自适应步长的Alpha-shape表面重建算法[J]. 数据采集与处理, 2019, 34(3): 491–499. doi: 10.16337/j.1004-9037.2019.03.012.

    LI Shilin and LI Hongjun. Surface reconstruction algorithm using self-adaptive step Alpha-shape[J]. Journal of Data Acquisition and Processing, 2019, 34(3): 491–499. doi: 10.16337/j.1004-9037.2019.03.012.
    [36] LÜTKEPOHL H. Vector Autoregressive Models[M]. HASHIMZADE N and THORNTON M A. Handbook of Research Methods and Applications in Empirical Macroeconomics. Cheltenham: Edward Elgar, 2013: 139–164. doi: 10.4337/9780857931023.00012.
    [37] GHIASI G, LIN T Y, and LE Q V. DropBlock: A regularization method for convolutional networks[C]. The 32nd International Conference on Neural Information Processing Systems, Montréal, Canada, 2018: 10750–10760.
    [38] BAI Shaojie, KOLTER J Z, and KOLTUN V. An empirical evaluation of generic convolutional and recurrent networks for sequence modeling[EB/OL]. https://arxiv.org/abs/1803.01271, 2018.
    [39] LUO Yi and MESGARANI N. Conv-TasNet: Surpassing ideal time-frequency magnitude masking for speech separation[J]. IEEE/ACM Transactions on Audio, Speech, and Language Processing, 2019, 27(8): 1256–1266. doi: 10.1109/TASLP.2019.2915167.
    [40] BADRINARAYANAN V, KENDALL A, and CIPOLLA R. SegNet: A deep convolutional encoder-decoder architecture for image segmentation[J]. IEEE Transactions on Pattern Analysis and Machine Intelligence, 2017, 39(12): 2481–2495. doi: 10.1109/TPAMI.2016.2644615.
    [41] CHO K, VAN MERRIËNBOER B, BAHDANAU D, et al. On the properties of neural machine translation: Encoder-decoder approaches[C]. Eighth Workshop on Syntax, Semantics and Structure in Statistical Translation, Doha, Qatar, 2014: 103–111. doi: 10.3115/v1/W14-4012.
    [42] VASWANI A, SHAZEER N, PARMAR N, et al. Attention is all you need[C]. The 31st International Conference on Neural Information Processing Systems, Long Beach, USA, 2017: 6000–6010.
    [43] SUBAKAN C, RAVANELLI M, CORNELL S, et al. Attention is all you need in speech separation[C]. ICASSP 2021–2021 IEEE International Conference on Acoustics, Speech and Signal Processing (ICASSP), Toronto, Canada, 2021: 21–25. doi: 10.1109/ICASSP39728.2021.9413901.
    [44] BARRON J T. A general and adaptive robust loss function[C]. 2019 IEEE/CVF Conference on Computer Vision and Pattern Recognition, Long Beach, USA, 2019: 4326–4334. doi: 10.1109/CVPR.2019.00446.
    [45] ALT H and GODAU M. Computing the Fréchet distance between two polygonal curves[J]. International Journal of Computational Geometry & Applications, 1995, 5(1/2): 75–91. doi: 10.1142/S0218195995000064.
  • 加载中
图(20) / 表(2)
计量
  • 文章访问数:  109
  • HTML全文浏览量:  66
  • PDF下载量:  41
  • 被引次数: 0
出版历程
  • 收稿日期:  2024-02-29
  • 修回日期:  2024-04-10
  • 网络出版日期:  2024-05-10

目录

    AI智能化辅读

    AI智能问答

        /

        返回文章
        返回