多扩展目标跟踪及优化处理方法研究

With the improvement of the resolution of advanced sensors such as modern radar, the abundant information of target is obtained, then the traditional point target has gradually evolved into the extended target, which causes the traditional point target tracking method not to meet the requirements of modern weapon system. Therefore, the research on extended-target tracking methods has become a hot issue of target tracking. Based on theories of random finite set (RFS), random matrix, probability graph model, this project is focused on the following research contents: (1) For modeling extended target shape, the efficient modeling method based on random matrix will be researched; (2) For tracking the extended target, the fast extended target Poisson multi-Bernoulli mixture (PMBM) filtering framework will be researched in the complex surveillance; (3) For optimizing the extended target tracking, joint detection and tracking, joint tracking and classification and track management based on RFS will be researched; (4) For the tracking performance evaluation, integrated tracking performance evaluation criteria related to various states of extended targets will be researched. The research of this project is of great significance both in theory and application, which perfects the extended target tracking system and optimization processing theory. This research can provide the corresponding theoretical and technical support for improving the performance of modern tracking system.

随着现代雷达等先进传感器分辨能力的不断提高，可获得的目标信息愈加丰富，目标描述由传统的点目标逐渐演变为扩展目标，致使传统点目标跟踪方法不能满足现代武器系统的要求，因此，研究适用于扩展目标的跟踪方法已成为目标跟踪领域的热点问题。本项目以随机有限集（RFS）、随机矩阵、概率图模型等理论为基础，重点开展以下研究内容：（1）在扩展目标形状模型方面，研究基于随机矩阵的高效扩展目标建模方法；（2）在扩展目标跟踪方面，研究复杂环境下快速扩展目标泊松多伯努利混合（PMBM）滤波框架；（3）在扩展目标优化处理方面，研究基于RFS的联合检测与跟踪、联合跟踪与分类、航迹管理等优化处理方法；（4）在跟踪性能评价方面，研究综合考虑扩展目标各个状态的跟踪性能评价准则。本项目研究对完善扩展目标跟踪体系以及优化处理理论具有重要的理论意义和应用价值，可为提高现代跟踪系统的性能提供相应的理论和技术支撑。

随着现代雷达等先进传感器分辨能力的不断提高，可获得的目标信息愈加丰富，目标描述由传统的点目标逐渐演变为扩展目标，致使传统点目标跟踪方法不能满足现代武器系统的要求，因此，研究适用于扩展目标的跟踪方法已成为目标跟踪领域的热点问题。本项目以随机有限集（Random Finite Set, RFS）、随机矩阵、概率图模型等新理论为基础，重点开展了扩展目标形状建模、多扩展目标跟踪、扩展目标优化处理、扩展目标跟踪性能评估等四个方面的研究，取得了一系列创新性研究成果，主要包括：（1）针对扩展目标形状建模问题，提出了基于量测变换的非线性滤波算法和子椭圆数目可变的非椭圆扩展目标跟踪算法，解决了扩展目标状态非线性估计和复杂形状估计问题；（2）针对多扩展目标跟踪问题，提出了基于联合似然的广义标签多伯努利（Generalized Labelled Multi-Bernoulli, GLMB）滤波算法和基于航迹/混合概率图模型的多扩展目标数据关联算法，解决了不同形状的多扩展目标跟踪问题；（3）针对扩展目标优化处理问题，提出了概率假设密度（Probability Hypothesis Density, PHD）扩展目标联合跟踪与分类算法、基于Gamma箱粒子和标签随机有限集（RFS）的扩展目标跟踪算法，解决了扩展目标优化和实时性问题；（4）针对扩展目标跟踪性能评估问题，构建了复杂形状扩展目标跟踪性能评价体系。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。