粒子滤波连续帧图像中弱目标检测前跟踪机理研究.pdfVIP

15 2 Vol. 15, No. 2 2010 2 Journal of mI age and Graphics Feb. 2010 龚亚信 杨宏文 胡卫东 郁文贤 ( A R , 410073) , Bayes, Bayes, , : N911. 73 : A : 1006-8961( 2010) 02-0307-07 Study onMechanism of Particle FilterBased Track-before-detect forDim Targets in SequentialFrame Images GONG Ya-xin, YANG Hong-wen, HU We-i dong, YU Wen-xian (ATR StateK ey Laboratory, N ational Un iv ersity of Def ense Technology, Changsha 410073) Abstract Particle filter based track-before-detect (PF- BD) is an active research topic on detection and track ing for dmi targets in sequential frame mi ages currently. Whereas there is absent of study on m echan ism of PF- BD in existingw ork, the princip le of BD which is based on cum ulative m easurem ent likelihood ratio to detect dmi targets is analyzed in Bayesian inference framework. he course of detection and tracking of dmi targets using Bayesian BD which is based on recursively cumu latem easurem ents is educed and the m echanism wh ich realizes it by unnorm alized particlesc w eights in particle filter processing is analyzed and confirmed in theory. A PF- BD algorithm is presented based on this m echan ism. Smi ulation resu lts show that the analysis is valid and PF- BD algorithm in th is paper is efficient. Key ords Bayesian inference, track-before-detect ( BD), particle filter, dmi targets , 0 , BD , ( BD) , , [ 1]