机动目标建模及跟踪方法研究

发布时间：2018-04-27 22:47

本文选题：机动目标跟踪 + 非线性滤波　；参考：《电子科技大学》2015年硕士论文

【摘要】：随着科学技术的迅速发展和现代战争战场环境日趋复杂,现代目标的机动性能越来越复杂多变,而人们对跟踪的性能要求则越来越高,因而,使得现代机动目标跟踪技术无论是在军事上还是民用上的都是一个研究热点。在机动目标跟踪领域中,非线性滤波技术是机动目标跟踪的主要工具,机动目标运动模型的建立是对机动目标实施跟踪的基本要素,而模型的结构性是实施机动目标跟踪的重要手段,因此,本文从以上三个方面对机动目标跟踪技术进行了研究。首先本文介绍了机动目标跟踪的基本原理,分析了当代跟踪系统中非线性因素的存在。并对扩展卡尔曼滤波算法、无迹卡尔曼滤波算法以及容积卡尔曼滤波算法进行了理论分析。然后通过对机动目标跟踪的仿真,分析了三种非线性滤波算法的优劣。并在后续的仿真中对这三种算法进行了应用。接着,本文介绍了几种机动目标运动数学模型,并研究了当前统计模型以及Jerk模型的主要原理。通过实验仿真,分析了当前统计模型以及Jerk模型对机动目标跟踪的效果。并重点分析了当前统计模型对非机动段跟踪精度不足的问题。在此基础上研究了基于模糊自适应的当前统计模型,与原有的当前统计模型进行了仿真对比,验证了改进的当前统计模型在机动段以及非机动段较原有当前统计模型对机动目标都有更优跟踪精度的结论。对于模型结构性方面的研究,本文首先研究了固定结构多模型算法中应用最为广泛的交互式模型算法,并对模型数目不同的交互式模型算法进行了对比,由此引出了固定结构多模型算法的局限性。然后介绍了变结构多模型算法的基本原理并对当前最主要的四种变结构多模型算法进行了研究,包括模型群切换算法、可能模型集算法、自适应网格算法以及期望模式扩充算法。并对后三种算法与固定结构的交互式多模型算法进行了仿真对比。并在针对原有的期望模式扩充算法在机动段跟踪能力不足的问题进行了改进,改进了一种基于强跟踪滤波器的期望模式扩充算法,从而进一步提高了期望模式扩充算法在机动段的跟踪能力。并将期望模式扩充的思想引入到了可能模型集算法里面,改进了一种引入期望模式扩充思想的可能模型集算法。相比于固定结构的交互式多模型算法,该算法不仅提升了跟踪精度,同时采用了较少的模型,提升了算法的时效性。
[Abstract]:With the rapid development of science and technology and the increasingly complex battlefield environment of modern war, the maneuverability of modern targets is becoming more and more complex and changeable, and the performance of tracking is becoming more and more demanding. It makes the modern maneuvering target tracking technology a research hotspot both in military and civilian fields. In the field of maneuvering target tracking, nonlinear filtering technology is the main tool of maneuvering target tracking, and the establishment of maneuvering target motion model is the basic element of maneuvering target tracking. The structure of the model is an important means of maneuvering target tracking, so this paper studies the maneuvering target tracking technology from the above three aspects. Firstly, the basic principle of maneuvering target tracking is introduced, and the existence of nonlinear factors in modern tracking system is analyzed. The extended Kalman filtering algorithm, the unscented Kalman filter algorithm and the volumetric Kalman filter algorithm are analyzed theoretically. Then, through the simulation of maneuvering target tracking, the advantages and disadvantages of three nonlinear filtering algorithms are analyzed. The three algorithms are applied in the subsequent simulation. Then, several mathematical models of maneuvering target motion are introduced, and the main principles of current statistical model and Jerk model are studied. The effect of the current statistical model and Jerk model on maneuvering target tracking is analyzed by experimental simulation. The problem of insufficient tracking accuracy of the current statistical model for non-maneuvering segment is analyzed. On this basis, the current statistical model based on fuzzy adaptive is studied and compared with the current statistical model. The conclusion that the improved current statistical model has better tracking accuracy for maneuvering targets in both maneuvering and non-maneuvering segments is verified. For the structural aspects of the model, this paper first studies the most widely used interactive model algorithm in the fixed structure multi-model algorithm, and compares the interactive model algorithm with different number of models. The limitation of the fixed-structure multi-model algorithm is introduced. Then the basic principle of variable structure multi-model algorithm is introduced, and four main variable structure multi-model algorithms are studied, including model group switching algorithm, possible model set algorithm. Adaptive mesh algorithm and expected pattern extension algorithm. The last three algorithms are compared with the fixed-structure interactive multi-model algorithm. In order to solve the problem of insufficient tracking ability of the original expected mode expansion algorithm in maneuvering segment, a new expected mode extension algorithm based on strong tracking filter is improved. Thus, the tracking ability of the expected mode extension algorithm in the maneuvering segment is further improved. The idea of expected pattern extension is introduced into the algorithm of the possible model set, and a possible model set algorithm which introduces the idea of expected pattern expansion is improved. Compared with the fixed structure interactive multi-model algorithm, the algorithm not only improves the tracking accuracy, but also uses fewer models to improve the timeliness of the algorithm.
【学位授予单位】：电子科技大学
【学位级别】：硕士
【学位授予年份】：2015
【分类号】：TN713

【共引文献】