基于概率假设密度的多传感器多目标跟踪技术研究

发布时间：2019-06-17 19:31

【摘要】：多目标跟踪技术作为多传感器信息融合领域的一个研究热点,在军事与民用领域有着广泛的应用。传统多目标跟踪方法以经典概率论为基础,其核心为解决多目标数据关联问题,跟踪过程中易受目标个数未知、杂波密集、检测率低等复杂环境影响,导致数据关联问题复杂度增加和跟踪精度下降。近年来,基于随机有限集(Random Finite Set,RFS)的概率假设密度(Probability Hypothesis Density,PHD)滤波方法颇受关注。该方法利用RFS理论,能够将目标状态集合和传感器测量集合统一描述于一个概率假设密度空间中,有效的避免了传统跟踪算法中的数据关联问题。尽管如此,目前大多数基于随机有限集的多目标跟踪方法还是针对单传感器提出的。在复杂环境下,很难做到仅依靠单个传感器获取的信息进行稳定且准确的滤波估计,通常需要融合多个传感器的信息来达到跟踪要求。为此,本文针对高杂波率和低检测率下的多传感器多目标跟踪问题进行研究,主要工作和研究成果如下:1)针对高杂波环境下,单传感器应用PHD滤波器出现跟踪效果退化的问题,通过构建了分布式多传感器数据融合结构模型,提出了一种基于高斯混合PHD滤波的自适应多传感器数据融合算法。仿真结果表明,与单传感器相比,所提算法有效的提高了跟踪精度。2)针对在不同的杂波环境以及检测率下,常规航迹融合算法具有的局限性,限制了跟踪效果的提高。为此,构建带反馈的分布式多传感器数据融合结构模型,并提出了两种不同的多传感器PHD融合算法:极值融合算法和乘积融合算法。通过不同场景的仿真实验,验证了所提算法性能优于传统算法。3)将常规多目标跟踪拓展到多机动目标跟踪,引入交互多模型(Interacting Multiple Model Algorithm,IMM)算法,构建了一种用于多机动目标跟踪的多传感器IMM-GMPHD滤波算法,使其能够有效处理杂波环境中的多机动目标跟踪问题。仿真实验证明:当目标发生机动时,所提算法能得到更高的目标状态估计精度。
[Abstract]:As a research focus in the field of multi-sensor information fusion, multi-target tracking technology has a wide range of applications in military and civil fields. The traditional multi-target tracking method is based on the classical probability theory. The core of the traditional multi-target tracking method is to solve the problem of multi-target data association. The tracking process is easily affected by the complex environment such as unknown number of targets, dense clutters, low detection rate and so on, which leads to the increase of the complexity of the data association problem and the decrease of tracking accuracy. In recent years, probabilistic hypothetical density (Probability Hypothesis Density,PHD) filtering methods based on stochastic finite set (Random Finite Set,RFS) have attracted much attention. By using RFS theory, the target state set and the sensor measurement set can be described in a probability hypothetical density space, which effectively avoids the problem of data association in the traditional tracking algorithm. However, most of the multi-target tracking methods based on random finite sets are proposed for single sensor. In complex environment, it is difficult to rely only on the information obtained by a single sensor for stable and accurate filtering estimation. It is usually necessary to fuse the information of multiple sensors to meet the tracking requirements. In this paper, the problem of multi-sensor multi-target tracking with high hash rate and low detection rate is studied. The main work and research results are as follows: 1) aiming at the degradation of tracking effect of single-sensor application PHD filter in high clutter environment, an adaptive multi-sensor data fusion algorithm based on Gao Si hybrid PHD filter is proposed by constructing the distributed multi-sensor data fusion structure model. The simulation results show that compared with the single sensor, the proposed algorithm effectively improves the tracking accuracy. 2) aiming at the limitations of the conventional track fusion algorithm in different clutter environment and detection rate, the tracking effect is limited. In this paper, a distributed multi-sensor data fusion structure model with feedback is constructed, and two different multi-sensor PHD fusion algorithms, extreme value fusion algorithm and product fusion algorithm, are proposed. The simulation results of different scenarios show that the proposed algorithm is superior to the traditional algorithm. 3) the conventional multi-target tracking is extended to multi-maneuvering target tracking, and the interactive multi-model (Interacting Multiple Model Algorithm,IMM algorithm is introduced to construct a multi-sensor IMM-GMPHD filtering algorithm for multi-maneuvering target tracking, which can effectively deal with the multi-maneuvering target tracking problem in cluttered environment. The simulation results show that the proposed algorithm can obtain higher accuracy of target state estimation when the target maneuvers.
【学位授予单位】：杭州电子科技大学
【学位级别】：硕士
【学位授予年份】：2017
【分类号】：TP212

【参考文献】