复杂环境下非椭圆扩展目标跟踪方法研究

With the increasing improvement of radar, infrared, etc. sensor resolution, extended target tracking has become a key technology needed to solve both in military and civilian fields. However, faced with the influence of complex environments such as low probabilities of detection and recognition, high target missing rate, etc., and constrained by uncertainties like environmental clutter, number of targets and target shape, how to realize the unified and precise mathematical modeling of extended targets, and research on the corresponding tracking algorithms are key scientific issues of the current target tracking field to be solved. For the non-ellipsoidal extended target tracking problem, this project will focus on the following aspects: (1) Based on the random finite set theory, multiple extended-target modeling and tracking problems in complex environment will be researched; (2) Based on the fuzzy adaptive resonance theory (ART), introducing Bayesian theory, the problems of association event approximation and sub-target number estimation in linear filtering will be researched, respectively; (3) Based on the box particle filter, introducing the rectangular classification features of fuzzy ART, the problem of nonlinear filtering modeling will be researched. Research on the project has an important role for improving the extended-target modeling and corresponding tracking theory, and provides the corresponding theoretical and technical supports for the non-ellipsoidal extended targets in complex environment.

随着雷达、红外等传感器分辨率的不断提高，扩展目标跟踪已成为军事及民用领域中亟需解决的关键技术。然而，面对低检测率、低识别率和高丢失率等复杂环境的影响，以及受环境杂波、目标数目、目标形状等不确定因素的制约，如何对扩展目标进行统一而精确的数学建模并研究相应的跟踪算法是当前目标跟踪领域亟待解决的关键科学问题。本项目针对非椭圆扩展目标跟踪问题，重点开展以下研究：（1）以随机有限集理论为基础，研究复杂环境下多扩展目标的建模与跟踪问题；（2）以模糊自适应谐振理论为基础，引入贝叶斯理论，研究线性滤波中存在的关联事件近似和子目标数估计问题；（3）以箱粒子滤波为基础，引入模糊自适应谐振理论中的矩形分类特性，研究非线性滤波建模问题。本项目研究对完善扩展目标建模与相应的跟踪理论具有重要的作用，可为复杂环境下非椭圆扩展目标的跟踪提供相应的理论和技术支撑。

随着雷达、红外等传感器分辨率的不断提高，扩展目标跟踪已成为军事及民用领域中亟需解决的关键技术。然而，面对低检测率、低识别率和高丢失率等复杂环境的影响，以及受环境杂波、目标数目、目标形状等不确定因素的制约，如何对扩展目标进行统一而精确的数学建模并研究相应的跟踪算法是当前目标跟踪领域亟待解决的关键科学问题。本项目针对非椭圆扩展目标跟踪问题，重点研究多扩展目标的建模与跟踪、线性滤波中存在的关联事件近似和子目标数估计问题、非线性滤波建模等关键问题，取得了一系列创新性研究成果，主要包括：（1）针对多扩展目标的建模与跟踪问题，提出了基于模糊ART的量测划分方法，从而提高了跟踪系统的效率；（2）针对线性滤波中存在的关联事件近似和子目标数估计问题，提出了子椭圆数目可变的机动多椭圆扩展目标建模方法和扩展目标联合跟踪与分类PHD滤波，可有效提高非椭圆扩展目标的近似度和子目标数估计的精度，便于系统后期的分类和识别；（3）针对非线性滤波建模问题，结合粒子滤波与箱粒子滤波方法，提出了基于箱粒子的扩展目标随机有限集非线性滤波方法，在提高系统跟踪精度的同时大大降低了系统的整体运行时间。本项目圆满完成了项目申请书既定的研究内容，研究成果在武器装备和防御系统的推广应用方面具有重要的价值。