粒子濾波技術(shù)通過(guò)非參數(shù)化的蒙特卡羅模擬方法實(shí)現(xiàn)遞推貝葉斯濾波，適用于非線性
目標(biāo)運(yùn)動(dòng)模型、非高斯噪聲的目標(biāo)跟蹤。在實(shí)際應(yīng)用中，初始粒子選取的精度對(duì)濾波效果往往會(huì)造成很大的影響；本文將競(jìng)爭(zhēng)策略應(yīng)用于初始粒子的選取，通過(guò)初始觀測(cè)值生成若干粒子群，并在濾波初始階段統(tǒng)計(jì)各粒子群中的有效粒子數(shù)量，最后選擇有效粒子數(shù)量最多的粒子群構(gòu)造初始化粒子。仿真結(jié)果表明，本算法可以提高初始粒子選取的精度，從而改善濾波效果。
關(guān)鍵詞：粒子濾波；初始化；目標(biāo)跟蹤
Abstract：Particle filter realizes recursive Bayes filter through non-parameter Monte Carlo
technique, it can be applied to target tracking in nonlinear and non-Gaussian environment. In
practice, the accuracy of initial particles will affect the result of filter heavily. In this paper, initial particles comes out through competition-cooperation strategy, that is, initial measurements give birth to several particle groups, each of the groups is tested separately in the beginning, the group with the most number of efficient particles is selected as the initial particles. Results of simulation show that the method improves the accuracy of initial particles, which makes the effect of filter much better.
Key words: Particle Filter; Initiation; Target Tracking