针对突变运动下的目标跟踪问题，提出一种基于视觉显著性的粒子滤波跟踪算法． 该算

法将基于视觉注意机制的视觉显著图引入粒子滤波框架中，根据视觉显著图的显著性区域，按

胜者为王和返回抑制机制进行目标检测-Aimingat solvingthetrackingproblemsunder thecircumstancesof abrupt motion，aparticlefilter trackerisproposedbasedonvisual saliencymodel． Thistrackerdetectsobject fromthesalient regionsinthesaliencymapby theway of winner-take-all andinhibition-of-return． Thedetectingresult istakenasaglobal proposal distributionandthenparticlesaresampledfromit　therefore，

theglobal statespacecanbesearchedinorder toavoidsuffering fromthelocal minimumproblem．

Moreover，inorder toincreasethesaliencyof theobject regioninthesaliency map，thebottom-up

andtop-downcomputational modelsarecombinedtogether． Thentheweightsof thefeaturemapsare

calculatedaccordingto thetarget templateandsaliency maps arefusedadaptively． Comparedwith

several other trackingalgorithms，