基于多假设序贯检验的分布式最速联合目标识别与跟踪

Quick recognition and accurate tracking of a target is prerequisite to battlefield surveillance, early warning and counterattack, and is of significance to national defense safety. Existing work on the theory and application of joint target recognition and tracking only try to increase the accuracy of recognition and tracking, and neglect the requirements that such applications actually need to recognize the target and start a track as quick as possible while the accuracy of recognition and tracking is guaranteed. To tackle this problem, this project first utilize the information contained in kinematic measurements to improve the target recognition and study the tracking-aided target quickest recognition method. Then, based-on the decision-performance guaranteed joint decision and estimation, this project study the sequential joint quickest decision and estimation, and its distributed algorithms with different fusion architectures, in the framework of sequential test to implement the quickest recognition and tracking by using measurements from multi-heterogeneous sensors. Finally, all methods proposed in this project will be evaluated by software and semi-hardware simulations, respectively. Distributed sequential joint decision and estimation, as the key study of this project, is an extension and intensification of distributed estimation fusion and distributed decision fusion. This study has practical values as well as the potential to further develop and enrich the theory and methods of information fusion.

快速的目标识别与精准的目标跟踪是战场监视、预警以及应对反击的前提条件，对我国国防安全具有重要的现实意义。现有的针对目标联合识别与跟踪的理论与应用研究，仅以提高识别率和跟踪精度为目的，忽略了实际应用要求在保证识别率和跟踪精度的条件下最快识别目标并生成航迹的迫切需求。为此，本项目首先利用目标航迹中所包含的类别信息提高目标识别的速度，进而研究跟踪辅助的最速目标识别；然后，在建立保证决策性能的联合估计与决策方法的基础上，以多假设序贯检验为基本框架，研究序贯联合决策与估计方法及其不同构架下的分布式算法，以实现利用多异类传感器对目标进行最速联合识别与跟踪；最后，通过数值仿真和半实物仿真验证本项目所提出方法的有效性。本项目重点研究的分布式序贯联合决策与估计，是对估计融合及决策融合理论的扩展和深化，其成果不仅具有重要的实际应用价值，还将进一步丰富和发展信息融合理论与方法。

快速的目标识别与精准的目标跟踪是战场监视、预警以及应对反击的前提条件，对我国国防安全具有重要的现实意义。现有的针对目标联合识别与跟踪的理论与应用研究，仅以提高识别率和跟踪精度为目的，忽略了实际应用要求在保证识别率和跟踪精度的条件下最快识别目标并生成航迹的迫切需求。为此，本项目利用目标航迹中所包含的类别信息提高目标识别的速度，进而研究跟踪辅助的最速目标识别，在建立保证决策性能的联合估计与决策方法的基础上，以多假设序贯检验为基本框架，研究序贯联合决策与估计方法及其不同构架下的分布式算法，以实现利用多异类传感器对目标进行最速联合识别与跟踪；对于多动平台的网络化跟踪问题，本项目基于多种通信机制，提出了具有不同特性的分布式协同跟踪方法；本项目还提出了一种改进的多动平台的协同定位算法、基于变结构多模型的航迹平滑算法、以及多种估计性能的综合评估指标。本项目重点研究的最速目标识别和网络化目标跟踪算法，是对估计融合及决策融合理论的扩展和深化。本项目所研究的算法可以广泛的应用于基于多异类传感器的目标识别与跟踪，其中，协同定位与网络化跟踪可以直接应用于卫星拒止环境下的协同定位、识别和跟踪。