拟蒙特卡罗粒子滤波算法（quasi-Monte-Carlo particle filter, QMC-PF）精度不高，运算复杂度大，难以满足雷达机动目标跟踪精确性和实时性需求. 为此，提出一种基于BP神经网络的新型拟蒙特卡罗粒子滤波算法. 该算法将大权重粒子通过QMC分裂采样产生低差异性的子代粒子，以此来替代低权重粒子，保证了样本的有
效性和多样性；同时利用BP神经网络计算子代粒子的权重，提高了滤波的精度和速度；最后在不同的模型中进行仿真. 实验结果表明，与QMC-PF相比，所提出的算法提高了精度和运算速度，适用于雷达机动目标的跟踪.
To address the difficulties in meeting the needs of precise and real-time radar maneuvering target tracking due to low precision and high computation complexity of quasi-Monte-Carlo particle filter (QMCPF), a new quasi-Monte-Carlo particle filter algorithm base on BP neural network (NQMC-PF) is proposed.
Through QMC fission sampling, this algorithm generates low-discrepancy progeny particles to replace the low-weight particles to guarantee validity and diversity of the samples. Meanwhile, the algorithm uses BP neural network to calculate the weight of offspring of particles. With different models, the algorithm is tested. Experimental results show that, compared to QMC-PF, the proposed algorithm can enhance precision and increase calculation speed, and thus is applicable to radar for tracking maneuvering targets.
[1]ARULAMPALAM M S, MASKELL S, GORDON N, CLAPP T. A tutorial on particle ?lters for on-line nonlinear/non-Gaussian Bayesian tracking [J]. IEEE Transactions on Signal Processing, 2002, 50(2): 174−188.
[2]栾海妍，江桦，刘小宝. 利用粒子滤波与支持向量机的数字混合信号单通道盲分离 [J]. 应用科学学报，2011, 29 (2): 195-202
LUAN Haiyan, JIANG Hua, LIU Xiaobao. Single channel blind source separation of digital mixtures using particle filtering and support vector machine [J]. Journal of Applied Sciences, 2011, 29 (2): 195-202. (in Chinese)
[3]POCOCK J A, DANCE S L, LAWLESS A S. State estimation using the particle filter with mode tracking [J]. Computers & Fluids Fluids, 2011,46(1): 392-397
[4]DOUCET A, GODSILL S. On sequential Monte Carlo sampling methods for Bayesian filtering [J]. Statistics and Computing, 2000,10(1): 197-208.
[5]YU Yihua, ZHENG Xuanyuan. Particle filter with ant colony optimization for frequency offset estimation in OFDM systems with unknown noise distribution [J]. Signal Processing. 2011,91(5): 1339-1342.
[6]KHAN Z, BALCH T, DELLAERT F. MCMC-based particle ?ltering for tracking a variable number of interacting targets [C]//IEEE Transactions on Pattern Analysis and Machine Intelligence, 2005, 27(11): 1805-1819
[7]ECUYER P. Quasi-Monte Carlo methods with applications in ?nance [J]. Finance and Stochastics, 2009, 13(3): 307-349
[8]JOE S, KUO F Y. Constructing Sobol sequences with better two-dimensional projections [J]. SIAM Journal on Scienti?c Computing, 2008, 30(5): 2635-2654
[9]GRAHAM I G, KUO F Y, NUYENS D, Scheichla R, Sloanb I.H. Quasi-Monte Carlo methods for elliptic PDEs with random coefficients and applications[J]. Journal of Computational Physics, 2011, 230(10): 3668-3694
[10]Guo Dong，Wang Xiaodong. Quasi-Monte Carlo ?ltering in nonlinear dynamic systems [J]//IEEE Transactions on Signal Processing, 2006, 54(6): 2087-209
[11]王牧云，严壮志，葛俊杰. 医用内窥镜图像校正的BP神经网络方法 [J]. 应用科学学报，2009, 27(5): 480-484
WANG Muyun, YAN Zhuangzhi, GE Junjie. Distortion correction of medical endoscopic images using BP neural network [J]. Journal of Applied Sciences, 2009, 27(5): 480-484. (in Chinese)
[12]叶龙，王京玲，张勤. 遗传重采样粒子滤波器 [J]. 自动化学报，2007,33(8): 885-887
YE Long, WANG Jingling, ZHANG Qin. Genetic resampling particle filter [J]. Acta Automatica Sinica, 2007,33(8): 885-887
