空时偏差多传感器系统信息融合算法

发布时间：2018-02-21 02:11

本文关键词： 信息融合多维分配空时偏差滤波目标跟踪　出处：《哈尔滨工业大学》2017年硕士论文　论文类型：学位论文

【摘要】：在多传感器信息融合系统中,传感器观测数据来源存在不确定性及观测数据可能存在的未知系统偏差是影响信息融合性能的两个关键问题。解决观测源不确定问题,需要研究通用、有效的多传感器数据关联方法。在对未知系统偏差配准的研究中,目前多数算法研究的是空间偏差配准和时间对齐问题,未对时间偏差及同时存在空时偏差的问题进行研究;而实际应用中观测数据时戳存在时延,传感器之间会存在时间偏差,需配准传感器间的时间和空间偏差来保证融合效果。本文针对多传感器信息融合系统中的数据关联和空时偏差估计问题,重点进行以下三方面研究。第一,基于多维分配的数据关联方法。以多被动传感器多目标跟踪为例,研究基于多维分配的数据关联方法。该算法首先利用航迹的先验信息建立预选波门,只对各传感器落入航迹波门内的量测数据产生关联假设并对其分配代价函数,利用各传感器量测数据和航迹之间一一对应的关系构建约束条件,利用拉格朗日松弛算法对约束条件依次松弛得到二维分配问题,利用拍卖算法求得二维分配问题的分配组合,基于分配结果对约束条件依次实施,最终得到各传感器量测数据的关联结果。仿真实验表明该算法能兼顾时间复杂度和关联正确率。此外该算法可应用于其它类型多传感器多目标跟踪系统。第二,数据率已知多传感器系统空时偏差配准方法。以两个传感器为例,分析了传感器观测与目标状态及空时偏差的关系,提出了传感器空时偏差与目标状态联合估计模型。对传感器数据率相同和不同两种情况均进行了研究,提出了批处理和序贯处理两种传感器空时偏差与目标状态联合估计方案。仿真结果验证了所提传感器空时偏差与目标状态联合估计模型和方法的有效性,可以同时实现空时偏差配准和目标状态的融合估计;其中,序贯处理方案能取得更优效果。另外,针对传感空时偏差估计值收敛较慢问题,在均方误差最小准则下,对多个目标同一时刻的空时偏差估计进行加权融合,仿真结果表明,多目标加权融合可以获得更精确的空时偏差估计,收敛速度更快。第三,数据率未知多传感器系统空时偏差配准方法。以两个传感器为例,在仅知传感器观测数据时戳且时戳存在固定延时的情况下,建立了未知数据率传感器空时偏差与目标状态联合估计模型,提出了序贯处理传感器空时偏差与目标状态联合估计方法。仿真结果验证所提未知数据率目标状态与传感器空时偏差联合估计模型与方法能有效的实现空时偏差的配准和目标状态的融合估计。同时在均方根误差最小准则下,对多个目标同一时刻的空时偏差估计进行加权融合以解决空时偏差估计值收敛慢的问题。仿真结果表明多目标加权融合能显著的提高空时偏差的估计精度和收敛速度。
[Abstract]:In the multi-sensor information fusion system, the uncertainty of the source of the sensor observation data and the unknown system deviation that may exist in the observation data are two key problems affecting the performance of the information fusion. It is necessary to study general and effective multi-sensor data association methods. In the research of unknown system bias registration, most of the current algorithms focus on spatial bias registration and time alignment. The problem of time deviation and space-time deviation is not studied, while in practical application, there is time delay in the time stamp of observation data, and there will be time deviation between sensors. It is necessary to register the time and space deviation between sensors to ensure the fusion effect. In this paper, the data association and space-time deviation estimation in multi-sensor information fusion system are studied in the following three aspects. Data association method based on multi-dimension assignment. Taking multi-passive sensors and multi-target tracking as an example, the data association method based on multi-dimension assignment is studied. Only the correlation hypothesis is generated and the cost function is assigned to the measured data of each sensor falling into the track gate, and the constraint condition is constructed by using the one-to-one correspondence between the measured data and the track of each sensor. The Lagrangian relaxation algorithm is used to relax the constraint conditions in turn to obtain the two-dimensional assignment problem, and the auction algorithm is used to obtain the allocation combination of the two-dimensional assignment problem. Based on the allocation results, the constraint conditions are implemented sequentially. Finally, the correlation results of the measured data of each sensor are obtained. The simulation results show that the algorithm can take into account both the time complexity and the correlation accuracy. In addition, the algorithm can be applied to other types of multi-sensor and multi-target tracking systems. Secondly, The relationship between sensor observation and target state and space-time deviation is analyzed with two sensors as an example. A joint estimation model of sensor space-time deviation and target state is proposed. Two joint estimation schemes of sensor space-time deviation and target state are proposed for batch processing and sequential processing. The simulation results verify the validity of the proposed model and method for the joint estimation of sensor space-time deviation and target state. The fusion estimation of space-time deviation registration and target state can be realized at the same time, in which the sequential processing scheme can obtain better results. In addition, for the problem of slow convergence of sensor space-time deviation estimation, under the minimum mean square error criterion, The weighted fusion of space-time deviation estimation of multiple targets at the same time shows that the multi-objective weighted fusion can obtain more accurate estimation of space-time deviation, and the convergence speed is faster. Third, Space-time bias registration method for multi-sensor systems with unknown data rate. Taking two sensors as an example, when only the time stamp of sensor observation data is known and the time stamp has a fixed time delay, The joint estimation model of space-time deviation and target state of unknown data rate sensor is established. A joint estimation method of sensor space-time deviation and target state is proposed in this paper, and the simulation results show that the proposed joint estimation model and method of target state and sensor space-time deviation of unknown data rate can effectively realize space-time deviation. At the same time, under the RMS error minimization criterion, In order to solve the problem of slow convergence of space-time deviation estimates, the weighted fusion of multiple targets at the same time is carried out. The simulation results show that the multi-objective weighted fusion can significantly improve the estimation accuracy and convergence speed of space-time deviations.
【学位授予单位】：哈尔滨工业大学
【学位级别】：硕士
【学位授予年份】：2017
【分类号】：TP212;TP202

【参考文献】