摘要
现代医学诊断依赖于细胞运动状态的检查结果。为了准确获取活体细胞在一定时期内包括新生、分裂、消亡的生命活动的相关信息,为医学病理鉴定与研究提供准确可靠的实数据与定量分析结果,本文针对活体细胞追踪问题,将基于多伯努利滤波器(Multi-Bernoulli filter,MeMBer)的多目标追踪技术引入微观细胞领域的追踪中。本文将细胞个体模拟为椭圆形,对目标形态进行了估计。运用数学形态学对椭圆形的长轴、短轴、核心坐标、倾斜角度等形态特征与运动特征进行测定。本文基于多伯努利滤波器推导了一种细胞追踪算法,在分析目标观测似然函数的基础上,把利用观测似然函数对预测得到的目标状态当成量测信息进行更新,从而消除预测时带来的误差与杂波的干扰。该方法可应用于一般细胞运动状态下的活体细胞追踪。通过仿真实验验证了所得算法的有效性。
Many medical diagnostic results depend on the results of the examination of the state of motion of the cells. In order to accurately obtain the information of living cells activities including newborn,division and extinction in a certain period of time,and provide accurate and reliable real data and quantitative analysis results for medical pathology identification and research,this paper will focus on the problem of live cell tracking. The multi-target tracking technique based on Multi-Bernoulli filter(MeMBer) is introduced into the micro cell tracking. The innovation lies in the simulation of the individual cells into elliptical shapes,which are facilitated with the fitting of target shape. The shape parameters of a cell are uniquely determined by using mathematical morphology to measure features including the elliptical long axis,short axis,core coordinates,and tilt angle etc. The method can be applied to living cell tracking in general cell motion state. Finally,the effectiveness of the proposed algorithm is verified by simulation experiments.
引文
[1]COMANICIU D,MEER P. Mean shift:A robust approach toward feature space analysis[J]. IEEE Transactions on Pattern Analysis and M achine Intelligence,2002,24(5):603-619.
[2]SETHIAN J A. Level set methods and fast marching methods[J].Journal of Computing and Information Technology,1999,11(1):1-33.
[3]COMANICIU D,MEER P. Mean shift:A robust approach toward feature space analysis[J]. IEEE Transactions on Pattern Analysis and M achine Intelligence,2002,24(5):603-619.
[4] MAHLER R. Nonadditive probability,finite-set statistics,and information fusion[C]//Proceedings of the 34thIEEE Conference on Decision and Control. New Orleans,LA,USA:IEEE,1995:1947-1952.
[5]CLARK C E. The greatest of a finite set of random variables[J].Operations Research,1961,9(2):145-162.
[6] MAHLER R P S. Multitarget Bayes filtering via first-order multitarget moments[J]. IEEE Transactions on Aerospace and Electronic systems,2003,39(4):1152-1178.
[7]VO B N,MA W K. The Gaussian mixture probability hypothesis density filter[J]. IEEE Transactions on Signal Processing,2006,54(11):4091-4104.
[8]CLARK D E,PANTA K,VO B N. The GM-PHD filter multiple target tracker[C]//2006 9thInternational Conference on Information Fusion. Florence,Italy:IEEE,2007:1-9.
[9] SIDENBALADH H. Multi-target particle filtering for the probability hypothesis density[C]//Proceedings of the 2003 Sixth International Conference of Information Fusion. Cairns,Queensland,Australia:IEEE,2003:800-806.
[10]GOSTAR A K,HOSEINNEZHAD R,BAB-HADIASHAR A.M ulti-bernoulli sensor control for multi-target tracking[C]//IEEE Eighth International Conference on Intelligent Sensors,Sensor Netw orks and Information Processing. M elbourne:IEEE,2013:1-8.
[11]VO B T,VO B N,CANTONI A. The cardinality balanced multitarget multi-Bernoulli filter and its implementations[J]. IEEE Transactions on Signal Processing,2009,57(2):409-423.
[12]杜金瑞.基于多伯努利滤波器的多扩展目标跟踪方法研究[D].兰州:兰州理工大学,2018.
[13]王战.多目标多伯努利滤波跟踪算法研究[D].西安:西安工程大学,2016.
[14]王冬.基于多伯努利滤波视频多目标跟踪算法[D].无锡:江南大学,2017.
[15]邹其兵.多伯努利滤波器及其在检测前跟踪中的应用[D].西安:西安电子科技大学,2012.
[16]魏怡,徐华中.基于误差理论的椭圆识别快速算法[J].信息与电子工程,2005,3(4):249-252.