异类传感器多目标检测跟踪与识别随机集模型  

Random set models of dissimilar sensors for multi-target detection,tracking and recognition

作　　者：石绍应[1,2] 王小谟[1] 曹晨[1] 张靖[1] 汪先超[2] 

机构地区：[1]中国电子科学研究院,北京100041 [2]空军预警学院,湖北武汉430019

出　　处：《系统工程与电子技术》2016年第12期2685-2691,共7页Systems Engineering and Electronics

基　　金：总装预研项目(51307020103)资助课题

摘　　要：为在空中预警监视系统中实现多异类传感器多目标联合检测、跟踪与识别,在多目标检测、跟踪的随机有限集模型基础上,进行多异类传感器多目标联合检测、跟踪与识别的理论模型与处理框架研究。通过对目标的运动学状态与目标识别属性状态统一描述,把多目标状态建模为一个用随机有限集描述的全局状态。通过对运动学传感器与属性传感器模型分析,把各异类传感器建模为一个全局传感器,并把各传感器的测量建模为一个用随机有限集描述的全局测量。根据全局状态与全局测量模型,把异类传感器多目标联合检测、跟踪与识别过程描述为Bayes滤波过程,并给出了相应的多异类传感器多目标联合检测、跟踪与识别处理框架。通过仿真试验验证了理论模型与框架的有效性。In order to detect, track, and recognize multi-target jointly by fusion multiple dissimilar sensors in the airborne warning system, the theoretical models and processing framework for multi-target joint detec- tion, tracking and recognition of dissimilar sensors are studied based on the random finite set theory. By descri- bing the single target^s kinematics states and recognition attribute states unifiedly, the multi-target states are modeled as a global state that is described by the random finite set. By analyzing the models of a kinematic sen- sor and an attribute sensor, the dissimilar sensors are modeled as a global sensor, and the measurements of those dissimilar sensors are modeled as a global measurement. Based on the models of global state and global measurement, the process of multi-target detection, tracking and recognition of dissimilar sensors are described by Bayes filtering, and the structure of multi-target recognition of dissimilar sensors fusion is established. Simu- lation results suggest that the proposed models and processing framework are executable and effective.

关 键 词：随机有限集 多目标联合检测 跟踪与识别 多异类传感器 融合 

分 类 号：TP391[自动化与计算机技术—计算机应用技术] TN911.7[自动化与计算机技术—计算机科学与技术]