摘要
对粒子滤波理论及其实现方法进行了研究.通过模拟实验验证了其优于卡尔曼跟踪的性能,并结合基于双正交小波的边缘形心提取方法和粒子滤波跟踪方法,构建了其跟踪框架.通过粒子数和系统状态转移方程的恰当选择,实现了云层背景下对背景简单的点目标和存在遮挡和旋转变化情况下的大目标进行跟踪.最后通过实验分析了粒子数目和状态方程的选取对跟踪精度的影响.实验证明,结合鲁棒性的小波检测方法和具有"多峰"描述的粒子滤波算法构造成的跟踪器,在运动目标存在局部遮挡和旋转变化等情况下能够实现稳定的目标跟踪.
Particle filter realize recursive Bayesian filter via Monte Carlo simulation.The method is suitable for any mon-linear system that could be represented with state model.The theory and method of particle filter are studied in the paper,through simulate experiment validate its capability excel to kalman tracking.A method of biorthogonal wavelet edge extraction combining particle filter is presented,and making up its frame of tracking.Through selecting state equation of system and particle number,which realize dot target in simple background and target in shelter complicated tracking used algorithm.At last the problems have analyzed whether the number of particles and state equation affects the tracking accuracy.Experiment validate,the tracker with combining robust wavelet inspect method and having many apex descriptive particle filter algorithm,which can realize steady target tracking in moving target exist local shelter.
出处
《东北师大学报(自然科学版)》
CAS
CSCD
北大核心
2010年第1期41-46,共6页
Journal of Northeast Normal University(Natural Science Edition)
基金
吉林省科技发展计划项目(20070322)
东北师范大学自然科学青年基金资助项目(20081003)