多伯努利滤波的快速红外弱小目标检测与跟踪  

Fast IR dim small targets detection and tracking based on the multi-Bernoulli filter

作　　者：李翠芸[1] 李宁[1] 姬红兵[1] 

机构地区：[1]西安电子科技大学电子工程学院,陕西西安710071

出　　处：《西安电子科技大学学报》2016年第4期69-74,190,共7页Journal of Xidian University

基　　金：国家自然科学基金资助项目(61372003);国家自然科学基金青年基金资助项目(61301289)

摘　　要：针对复杂背景下多个弱目标检测与跟踪中存在的跟踪不稳定、非实时及量测模型高度非线性问题,提出一种基于多伯努利滤波的快速检测与跟踪算法.首先,采用改进的Robinson Guard算法抑制背景杂波,避免强起伏背景图像中目标被抑制的问题;其次,采用平方根容积卡尔曼滤波实现多伯努利检测前跟踪,在保证实时跟踪的同时,解决了滤波的高度非线性,避免了协方差矩阵负定造成的数值不稳定.实测红外背景图像实验表明,改进的Robinson Guard算法能够有效抑制背景杂波、保留弱目标信息,平方根容积卡尔曼多伯努利检测前跟踪能更准、更稳定地估计目标数目和状态,实现目标的实时检测与跟踪.In view of the problems of tracking instability and non real-time and highly nonlinear measurement model for the detection and tracking of multiple Infrared （IR） dim targets under the complicated background, a fast detection and tracking algorithm based on multi-Bernoulli filter is presented. Firstly, the improved Robinson Guard filter suppresses the background clutter in order to avoid the suppression of the target in the strong background image. Then, the square-root cubature Kalman filter （SCK） performs the multi-Bernoulli track-before-detect （MB-TBD）. It can perform real-time tracking in the highly nonlinear model. The numerical instability caused by a negative definite covariance matrix is avoided. The experiments with some real IR background images show that the improved Robinson Guard algorithm can suppress the background clutter effectively and preserve the dim target information. It is also shown that SCK-MB-TBD can estimate the number and states of targets accurately and stably so as to complete target detection and tracking.

关 键 词：红外图像序列 背景杂波抑制 ROBINSON Guard滤波 平方根容积卡尔曼 多伯努利滤波 检测前跟踪 

分 类 号：TN953[电子电信—信号与信息处理]