粒子滤波多特征融合算法跟踪精度估计

Tracking Accuracy Estimation of Particle Filter Multi-Cue Fusion Algorithm

为了对粒子滤波的多特征融合算法的跟踪精度进行估计,本文提出一种基于密度函数小波估计的跟踪精度估计方法.利用正交对称小波基函数对各特征距离值集进行处理.根据跟踪过程中各特征的不同状态,对各特征距离值集的概率分布进行融合.由总距离值密度函数求出该帧的粒子权值期望,作为跟踪精度值,并从中估计出各特征距离值集的概率分布;形成反映总的跟踪状态的距离值密度函数.

In order to estimate the tracking accuracy of the particle filter multi-cue fusion algorithm ,a method based on the density functional wavelet estimation is presented .With the symmetric orthogonal wavelet function ,the distance values of each feature set are processed .According to the different states of each feature in the process of tracking ,the probability distribution of the distance values of each feature is integrated .Finally ,the desired particle weight of the frame is obtained as the tracking accuracy by the total distance density value function .The probability distributions of the distance value set of each feature are estimated ,from which the total distance value density function of the tracking state is formed .