基于区间分析技术的随机集多扩展目标跟踪关键技术研究

With the continuous improvement of radar and infrared sensor resolution, extended target tracking problem has become a hot spot in research of multiple target tracking field . This project aimed at multiple extended target tracking problem of high dimension, multi-scale, uncertainty, and the non-gaussian development direction, so as to improve multiple extended target detection and tracking ability in the complex environment as the goal, combining interval analysis theory, and researching on the key technology in the extension target tracking problem ,which based on interval particle random set; Through research and analysis of the extended target measurement model and measurement interval representation, structure measurement likelihood function model that is suitable for the interval measurement; At the same time, combining the advantage of interval analysis technology, put forward a measurement set partitioning method with the ability of automatic processing extended target fission and fusion; Based on this, define uniform distribution similarity measure criterion and the similarity distribution merger criterion, and extract the state of extended target accurately. This research results produced by the project will provide more theory and technology support to multiple extended targets application under complex environment, which has certain practical significance to improve situational awareness under complex environment, thus to achieve information superiority and decision superiority.

随着雷达、红外等传感器分辨率的不断提高，扩展目标跟踪问题已成为多目标跟踪领域研究的一个热点。本项目瞄准多扩展目标跟踪问题呈现高维数、多尺度、不确定、非高斯的发展方向，以提高复杂环境下多扩展目标探测和跟踪能力为目标，结合区间分析理论，针对基于区间粒子随机集多扩展目标跟踪问题的关键技术展开研究；通过研究和分析扩展目标的量测模型和量测的区间表示方法，构造适用于区间量测的量测似然函数模型；同时，结合区间分析技术的优势，提出具有自动处理扩展目标分裂和融合的量测集划分方法；在此基础上，定义均匀分布相似性度量准则和相似性分布合并准则，对扩展目标的状态进行精确提取。本项目研究所产生的研究成果必将为复杂环境下多扩展目标的应用提供理论和技术支撑，对复杂环境下态势感知能力的提高，从而获取信息优势乃至决策优势均有一定的现实意义。

面向多扩展目标跟踪问题，基于区间粒子随机集方法，重点研究区间量测似然函数的建立与量测集的划分，构建未知杂波环境下多目标跟踪与扩展目标起始模型，并基于该模型给出具体跟踪算法。针对扩展目标区间量测似然函数模型的建立问题，利用星-凸随机超曲面模型对量测的扩散程度进行建模，在扩展目标高斯混合概率密度滤波的框架下，提出一种基于星-凸形随机超曲面模型的扩展目标高斯混合概率密度滤波算法。针对杂波环境下量测密度差别较大的多扩展目标量测集划分问题，引入近邻传播聚类技术，提出了一种新的量测集划分算法。在未知杂波环境中，传统的多目标概率假设密度滤波器跟踪精度无法保证，所需粒子支撑集过大导致效率低下的问题，引入了区间分析技术，提出了未知杂波状态下基于箱粒子滤波的鲁棒标签多伯努利跟踪算法；针对杂波环境下多扩展目标跟踪中量测集划分和航迹起始问题，基于最近邻指数法对量测集进行聚类趋势分析，提出了一种基于扩展目标混合高斯概率假设密度滤波器的跟踪算法。研究成果可进一步完善区间随机集方法的理论与技术、解决多扩展目标跟踪的实际应用问题，具有重要的理论价值与实际意义。