基于粒子滤波的检测前跟踪算法研究
发布时间：2019-06-07 14:51
【摘要】： 弱小目标的检测与跟踪是红外预警系统、精确制导系统、卫星遥感系统中的一项关键技术。在长距离衰减和强噪声影响下,传感器接收的目标信噪比极低,此时传统的目标检测与跟踪方法已很难满足要求。近年来出现的检测前跟踪(TBD)方法为解决此问题提供了一条有效途径,这种方法集检测和跟踪于一体,在充分利用未经阈值化处理的传感器数据的基础上,通过时间积累目标能量,从而提高信噪比,实现对弱小目标的检测与跟踪。 基于粒子滤波(PF)的TBD算法性能优越,但粒子滤波一般需要大量的随机样本才能保证其性能,而大量随机样本的预测、更新和重采样计算使得粒子滤波很难满足工程上的实时性要求。本文重点研究了基于粒子滤波的TBD算法,采用多种技术减轻算法的计算负担,提高算法的实时性。 首先,通过对红外弱小目标模型研究分析,提出一个基于边缘化粒子滤波的TBD算法。该算法的特点在于采用边缘化方法,把目标状态中具有线性高斯特征的目标速度状态分离出来,对其使用线性最优的卡尔曼滤波,而目标位置、强度等非线性状态则仍用粒子滤波处理。这样不仅降低了粒子滤波估计状态的维数,大大减少了计算量,而且还提高了算法在低信噪比下的检测性能和跟踪精度。 其次,用一种误差收敛更快的拟蒙特卡罗(QMC)积分替代粒子滤波中传统的蒙特卡罗(MC)积分方法,提出一个改进算法:基于拟蒙特卡罗的高斯粒子滤波(QMC-GPF)。由于QMC积分能用较少的、分布规整的样本点达到MC积分的精度,该算法能在保证精度的前提下节省了大量计算负担。 最后,在QMC-GPF算法的基础上,利用滤波状态协方差矩阵在跟踪过程中的收敛特性构建判断逻辑,实现目标检测。算法结构简单,计算量小,仿真实验和实测数据实验显示,该算法对3dB以上的目标具有良好的跟踪检测能力。
[Abstract]:The detection and tracking of weak and small targets is a key technology in infrared early warning system, precision guidance system and satellite remote sensing system. Under the influence of long distance attenuation and strong noise, the signal-to-noise ratio (SNR) of the target received by the sensor is very low, so the traditional target detection and tracking method is difficult to meet the requirements. In recent years, the pre-detection tracking (TBD) method provides an effective way to solve this problem. This method integrates detection and tracking, and makes full use of the sensor data without threshold processing. The energy of the target is accumulated by time, so as to improve the signal-to-noise ratio (SNR) and realize the detection and tracking of weak and small targets. The TBD algorithm based on particle filter (PF) has excellent performance, but particle filter generally needs a large number of random samples to ensure its performance, and a large number of random samples predict. Update and resampling calculation make it difficult for particle filter to meet the real-time requirements of engineering. In this paper, the TBD algorithm based on particle filter is studied. A variety of techniques are used to reduce the computational burden of the algorithm and improve the real-time performance of the algorithm. Firstly, through the research and analysis of infrared small and weak target model, a TBD algorithm based on marginalized particle filter is proposed. The characteristic of this algorithm is that the target speed state with linear Gao Si characteristics in the target state is separated by using the marginalization method, and the linear optimal Kalman filter is used for the target state, and the target position is used. The nonlinear states such as intensity are still treated by particle filter. This not only reduces the dimension of particle filter estimation state and greatly reduces the amount of computation, but also improves the detection performance and tracking accuracy of the algorithm at low signal-to-noise ratio (SNR). Secondly, a quasi-Monte Carlo (QMC) integral with faster error convergence is used to replace the traditional Monte Carlo (MC) integral method in particle filter, and an improved algorithm is proposed: Gao Si particle filter (QMC-GPF) based on quasi-Monte Carlo. Because the QMC integral can be used less and the distributed sample points can achieve the accuracy of MC integral, the algorithm can save a lot of computational burden on the premise of ensuring the accuracy. Finally, based on the QMC-GPF algorithm, the judgment logic is constructed by using the convergence characteristics of the filtered state covariance matrix in the tracking process, and the target detection is realized. The algorithm has the advantages of simple structure and small amount of computation. Simulation experiments and measured data experiments show that the algorithm has good tracking and detection ability for targets above 3dB.
【学位授予单位】：西安电子科技大学
【学位级别】：硕士
【学位授予年份】：2009
【分类号】：TP391.41
本文编号：2494880
[Abstract]:The detection and tracking of weak and small targets is a key technology in infrared early warning system, precision guidance system and satellite remote sensing system. Under the influence of long distance attenuation and strong noise, the signal-to-noise ratio (SNR) of the target received by the sensor is very low, so the traditional target detection and tracking method is difficult to meet the requirements. In recent years, the pre-detection tracking (TBD) method provides an effective way to solve this problem. This method integrates detection and tracking, and makes full use of the sensor data without threshold processing. The energy of the target is accumulated by time, so as to improve the signal-to-noise ratio (SNR) and realize the detection and tracking of weak and small targets. The TBD algorithm based on particle filter (PF) has excellent performance, but particle filter generally needs a large number of random samples to ensure its performance, and a large number of random samples predict. Update and resampling calculation make it difficult for particle filter to meet the real-time requirements of engineering. In this paper, the TBD algorithm based on particle filter is studied. A variety of techniques are used to reduce the computational burden of the algorithm and improve the real-time performance of the algorithm. Firstly, through the research and analysis of infrared small and weak target model, a TBD algorithm based on marginalized particle filter is proposed. The characteristic of this algorithm is that the target speed state with linear Gao Si characteristics in the target state is separated by using the marginalization method, and the linear optimal Kalman filter is used for the target state, and the target position is used. The nonlinear states such as intensity are still treated by particle filter. This not only reduces the dimension of particle filter estimation state and greatly reduces the amount of computation, but also improves the detection performance and tracking accuracy of the algorithm at low signal-to-noise ratio (SNR). Secondly, a quasi-Monte Carlo (QMC) integral with faster error convergence is used to replace the traditional Monte Carlo (MC) integral method in particle filter, and an improved algorithm is proposed: Gao Si particle filter (QMC-GPF) based on quasi-Monte Carlo. Because the QMC integral can be used less and the distributed sample points can achieve the accuracy of MC integral, the algorithm can save a lot of computational burden on the premise of ensuring the accuracy. Finally, based on the QMC-GPF algorithm, the judgment logic is constructed by using the convergence characteristics of the filtered state covariance matrix in the tracking process, and the target detection is realized. The algorithm has the advantages of simple structure and small amount of computation. Simulation experiments and measured data experiments show that the algorithm has good tracking and detection ability for targets above 3dB.
【学位授予单位】：西安电子科技大学
【学位级别】：硕士
【学位授予年份】：2009
【分类号】：TP391.41
【引证文献】
相关硕士学位论文 前7条
1 杨心力;基于粒子滤波的视频目标跟踪研究[D];湘潭大学;2011年
2 郭辉;基于非线性滤波的目标跟踪算法研究[D];西安电子科技大学;2010年
3 李倩;基于FPGA的非线性滤波算法实现研究[D];西安电子科技大学;2010年
4 郭姗姗;基于改进粒子滤波的红外弱小目标检测前跟踪算法[D];哈尔滨工程大学;2012年
5 邹其兵;多伯努利滤波器及其在检测前跟踪中的应用[D];西安电子科技大学;2012年
6 刘铮;自适应颜色直方图的粒子滤波算法[D];武汉理工大学;2012年
7 杨瑞兴;基于粒子滤波的雷达弱目标TBD算法研究[D];西安电子科技大学;2013年
，本文编号：2494880
本文链接：https://www.wllwen.com/kejilunwen/dianzigongchenglunwen/2494880.html
