Abstract: Track-to-track association (T2TA), which aims at unifying track batch numbers and reducing track redundancy, serves as a precondition and foundation for track fusion and situation awareness. The current problems of T2TA come mainly from two sources: track data and association methods. Ubiquitous problems include errors and inconsistent update periods in track data, as well as suboptimal association results and dependencies on prior information and assumed motion models for association methods. Focusing on these two aspects, we propose a multiple-hypothesis algorithm for multi-sensor T2TA with an intelligent track score (MH-T2TA). A spatial–temporal registration module is designed based on self-attention and a contrastive learning architecture to eliminate errors and unify the distributions of asynchronous tracks. A multiple-hypothesis algorithm is combined with deep learning to estimate the association score of a pair of tracks without relying on prior information or assumed motion models, and the optimal association pairs can be obtained. With three kinds of loss functions, tracks coming from the same targets become closer, tracks coming from different targets become more distant, and the estimated track scores are very similar to the real ones. Experimental results demonstrate that the proposed MH-T2TA can associate tracks in complex scenarios and outperform other T2TA methods.

MH-T2TA：基于智能航迹评分的多传感器航迹关联多假设算法

[1]Alt H, Godau M, 1995. Computing the Fréchet distance between two polygonal curves. Int J Comp Geom Appl, 5(01n02):75-91.

[2]Aziz AM, 2011. A new fuzzy clustering approach for data association and track fusion in multisensor-multitarget environment. IEEE Aerospace Conf, p.1-10.

[3]Ba JL, Kiros JR, Hinton GE, 2016. Layer normalization. https://arxiv.org/abs/1607.06450

[4]Bar-Shalom Y, 2008. On the sequential track correlation algorithm in a multisensor data fusion system. IEEE Trans Aerosp Electron Syst, 44(1):396.

[5]Bar-Shalom Y, Li XR, 1995. Multitarget-Multisensor Tracking: Principles and Techniques. YBS Publishing, Storrs, USA.

[6]Bar-Shalom Y, Fortmann TE, Cable PG, 1990. Tracking and data association. J Acoust Soc Am, 87(2):918-919.

[7]Blackman SS, 1986. Multiple-Target Tracking with Radar Applications. Artech House, Dedham, USA.

[8]Blackman SS, 2004. Multiple hypothesis tracking for multiple target tracking. IEEE Aerosp Electr Syst Mag, 19(1):5-18.

[9]Blackman SS, Dempster RJ, Broida TJ, 1993. Multiple hypothesis track confirmation for infrared surveillance systems. IEEE Trans Aerosp Electron Syst, 29(3):810-824.

[10]Blostein SD, Richardson HS, 1994. A sequential detection approach to target tracking. IEEE Trans Aerosp Electron Syst, 30(1):197-212.

[11]Chen L, Ng R, 2004. On the marriage of Lp-norms and edit distance. Proc Int Conf on Very Large Data Bases, p.792-803.

[12]Chen L, Özsu MT, Oria V, 2005. Robust and fast similarity search for moving object trajectories. Proc ACM SIGMOD Int Conf on Management of Data, p.491-502.

[13]Chong C, Mori S, Tse E, et al., 1982. Distributed Hypothesis Formation in Distributed Sensor Networks. Advanced Information and Decision Systems Report, No. IR-1 I015-l.

[14]Cox IJ, Hingorani SL, 1996. An efficient implementation of Reid’s multiple hypothesis tracking algorithm and its evaluation for the purpose of visual tracking. IEEE Trans Patt Anal Mach Intell, 18(2):138-150.

[15]Crouse DF, 2016. On implementing 2D rectangular assignment algorithms. IEEE Trans Aerosp Electron Syst, 52(4):1679-1696.

[16]Cui YQ, Liu Y, Tang TT, et al., 2021. A new adaptive track correlation method for multiple scenarios. IET Radar Sonar Navig, 15(9):1112-1124.

[17]Cui YQ, Xu PL, Gong C, et al., 2023. Multisource track association dataset based on the global AIS. J Electr Inform Technol, 45(2):746-756 (in Chinese).

[18]Devlin J, Chang MW, Lee K, et al., 2018. BERT: pre-training of deep bidirectional transformers for language understanding. https://arxiv.org/abs/1810.04805

[19]Du RZ, Liu L, Bai XR, et al., 2021. A new scatterer trajectory association method for ISAR image sequence utilizing multiple hypothesis tracking algorithm. IEEE Trans Geosci Remote Sens, 60:1-13.

[20]Du W, Ning HS, Wei Y, et al., 2013. Fuzzy double-threshold track association algorithm using adaptive threshold in distributed multisensor-multitarget tracking systems. IEEE Int Conf on Green Computing and Communications and IEEE Internet of Things and IEEE Cyber, Physical and Social Computing, p.1133-1137.

[21]Hadsell R, Chopra S, LeCun Y, 2006. Dimensionality reduction by learning an invariant mapping. IEEE Computer Society Conf on Computer Vision and Pattern Recognition, p.1735-1742.

[22]Hausdorff F, 1914. Grundzüge der Mengenlehre. Veit Comp., Leipzig, Germany.

[23]He Y, Zhang JW, 2006. New track correlation algorithms in a multisensor data fusion system. IEEE Trans Aerosp Electron Syst, 42(4):1359-1371.

[24]He Y, Wang GH, Guan X, 2010. Information Fusion Theory with Applications. Electronic Industry Press, Beijing, China (in Chinese).

[25]Hochreiter S, Schmidhuber J, 1997. Long short-term memory. Neur Comput, 9(8):1735-1780.

[26]Jin B, Tang YF, Zhang ZK, et al., 2023. Radar and AIS track association integrated track and scene features through deep learning. IEEE Sens J, 23(7):8001-8009.

[27]Kanyuck AJ, Singer RA, 1970. Correlation of multiple-site track data. IEEE Trans Aerosp Electron Syst, AES-6(2):180-187.

[28]Klein I, Bar-Shalom Y, 2016. Tracking with asynchronous passive multisensor systems. IEEE Trans Aerosp Electron Syst, 52(4):1769-1776.

[29]Kurien T, 1990. Issues in the design of practical multitarget tracking algorithms. In: Yaakov BS (Ed.), Multitarget-Multisensor Tracking: Advanced Applications. Artech House, Norwood, USA, p.43-84.

[30]Lee CS, Whang IH, Ra WS, 2023. Knowledge‐based multiple hypothesis tracking and identification of manoeuvring reentry targets. IET Radar Sonar Navig, 17(10):1479-1497.

[31]Loshchilov I, Hutter F, 2016. SGDR: stochastic gradient descent with warm restarts. https://arxiv.org/abs/1608.03983

[32]Loshchilov I, Hutter F, 2017. Decoupled weight decay regularization. https://arxiv.org/abs/1711.05101

[33]Misra D, 2019. Mish: a self regularized non-monotonic activation function. https://arxiv.org/abs/1908.08681

[34]Paszke A, Gross S, Massa F, et al., 2019. PyTorch: an imperative style, high-performance deep learning library. Proc 33^rd Conf on Neural Information Processing System, p.1-12.

[35]Qi L, Dong K, Liu Y, et al., 2017. Anti-bias track-to-track association algorithm based on distance detection. IET Radar Sonar Navig, 11(2):269-276.

[36]Qi L, He Y, Dong K, et al., 2018. Multi-radar anti-bias track association based on the reference topology feature. IET Radar Sonar Navig, 12(3):366-372.

[37]Reid DB, 1977. A Multiple Hypothesis Filter for Tracking Multiple Targets in a Cluttered Environment. Lockhead Missiles Space Company, Incorporated, USA.

[38]Reid DB, 1979. An algorithm for tracking multiple targets. IEEE Trans Autom Contr, 24(6):843-854.

[39]Shi Y, Wang Y, Wang SG, 2006. Fuzzy data association based on target topology of reference. J Nat Univ Def Technol, 28(4):105-109 (in Chinese).

[40]Sönmez HH, Hocaoğlu AK, 2022. Asynchronous track-to-track association algorithm based on reference topology feature. Sign Imag Video Process, 16(3):789-796.

[41]Su JL, Lu Y, Pan SF, et al., 2021. RoFormer: enhanced transformer with rotary position embedding. https://arxiv.org/abs/2104.09864

[42]Sun WF, Li XT, Pang ZZ, et al., 2023. Track-to-track association based on maximum likelihood estimation for T/R-R composite compact HFSWR. IEEE Trans Geosci Remote Sens, 61:5102012.

[43]Tian W, Wang Y, Shan XM, et al., 2014. Track-to-track association for biased data based on the reference topology feature. IEEE Signal Process Lett, 21(4):449-453.

[44]Tokta A, Hocaoglu AK, 2019. Sensor bias estimation for track-to-track association. IEEE Signal Process Lett, 26(10):1426-1430.

[45]Vaswani A, Shazeer N, Parmar N, et al., 2017. Attention is all you need. https://arxiv.org/abs/1706.03762

[46]Vlachos M, Kollios G, Gunopulos D, 2002. Discovering similar multidimensional trajectories. Proc 18^th Int Conf on Data Engineering, p.673-684.

[47]Wang J, Zeng YJ, Wei SM, et al., 2021. Multi-sensor track-to-track association and spatial registration algorithm under incomplete measurements. IEEE Trans Signal Process, 69:3337-3350.

[48]Werthmann JR, 1992. Step-by-step description of a computationally efficient version of multiple hypothesis tracking. Proc SPIE, Signal and Data Processing of Small Targets, 1698:288-300.

[49]Wu ZH, Pan SR, Chen FW, et al., 2021. A comprehensive survey on graph neural networks. IEEE Trans Neur Netw Learn Syst, 32(1):4-24.

[50]Xiong W, Xu PL, Cui YQ, et al., 2021. Track segment association via track graph representation learning. IET Radar Sonar Navig, 15(11):1458-1471.

[51]Xiong W, Xu PL, Cui YQ, 2024. Unsupervised and interpretable track‐to‐track association based on homography estimation of radar bias. IET Radar Sonar Navig, 18(2):294-307.

[52]Xu Z, Fang L, 2021. An improved track association algorithm based on AdaBoost and decision tree. IEEE 4^th Int Conf on Advanced Electronic Materials, Computers and Software Engineering, p.794-800.

[53]Yang YP, Yang F, Sun LG, et al., 2022. Multi-target association algorithm of AIS-radar tracks using graph matching-based deep neural network. Ocean Eng, 266:112208.

[54]Zhao HC, Sha ZC, Wu J, 2017. An improved fuzzy track association algorithm based on weight function. IEEE 2^nd Advanced Information Technology, Electronic and Automation Control Conf, p.1125-1128.

[55]Zhu HY, Chen S, 2014. Track fusion in the presence of sensor biases. IET Signal Process, 8(9):958-967.

[56]Zhu HY, Han SY, 2014. Track-to-track association based on structural similarity in the presence of sensor biases. J Appl Math, 2014:1-8.

[57]Zhu HY, Wang W, Wang C, 2016. Robust track-to-track association in the presence of sensor biases and missed detections. Inform Fus, 27:33-40.