本文關(guān)鍵詞：基于序貫貝葉斯濾波器的多目標(biāo)跟蹤方法研究　出處：《深圳大學(xué)》2017年碩士論文　論文類型：學(xué)位論文

  更多相關(guān)文章： 多目標(biāo)跟蹤 邊緣分布 存在概率 跳變馬爾可夫系統(tǒng)模型

【摘要】：多目標(biāo)跟蹤的主要目的是在目標(biāo)運(yùn)動模式具有不確定性、測量存在不確定性以及存在雜波等情況時,能夠有效地檢測出各個目標(biāo)并估計(jì)出各目標(biāo)的狀態(tài)。傳統(tǒng)的多目標(biāo)跟蹤方法通常使用數(shù)據(jù)關(guān)聯(lián)技術(shù),然而數(shù)據(jù)關(guān)聯(lián)可能會出現(xiàn)“組合爆炸”和計(jì)算量呈指數(shù)增長等問題。針對這一問題,Mahler提出了基于隨機(jī)有限集的概率假設(shè)密度(PHD),該濾波器不僅避免數(shù)據(jù)關(guān)聯(lián),而且解決了虛警、漏檢和目標(biāo)數(shù)未知情況下的多目標(biāo)跟蹤問題。雖然PHD濾波器在多目標(biāo)跟蹤過程中擁有許多的優(yōu)勢,但是該濾波器也存在著一些問題。首先,該濾波器很難將近距離的目標(biāo)區(qū)分開來。其次,該濾波器對所收到的測量數(shù)據(jù)進(jìn)行集中處理,如果數(shù)據(jù)處理不及時,則會導(dǎo)致后面數(shù)據(jù)處理延遲。最后,在低檢測概率情況下容易造成目標(biāo)信息丟失和目標(biāo)數(shù)估計(jì)不穩(wěn)定。針對上述的問題,我們提出了一種序貫多目標(biāo)貝葉斯濾波器。另外,為了使該濾波器適用于多機(jī)動目標(biāo)的跟蹤,我們提出了一種跳變馬爾可夫系統(tǒng)模型的序貫多目標(biāo)貝葉斯濾波器。論文的主要內(nèi)容如下:1)介紹了基于有限集統(tǒng)計(jì)學(xué)的多目標(biāo)貝葉斯濾波理論,討論了多目標(biāo)跟蹤模型,概述了最優(yōu)多目標(biāo)貝葉斯(Bayes)濾波器以及傳遞多目標(biāo)聯(lián)合后驗(yàn)分布一階矩的PHD濾波器。最后介紹了傳遞目標(biāo)邊緣分布和存在概率的邊緣分布貝葉斯(MDB)濾波器。2)研究并提出了一種序貫多目標(biāo)貝葉斯濾波器,該濾波器傳遞目標(biāo)的邊緣分布和存在概率并序貫處理當(dāng)前時刻收到的測量數(shù)據(jù)。同時,分別給出了適用于線性高斯系統(tǒng)和非線性高斯系統(tǒng)的序貫多目標(biāo)貝葉斯濾波器實(shí)現(xiàn)方法。仿真實(shí)驗(yàn)結(jié)果表明,在存在雜波、漏檢、目標(biāo)數(shù)目未知的情況時,該濾波器具有更好的多目標(biāo)跟蹤能力。3)為了解決多機(jī)動目標(biāo)的跟蹤問題,我們將跳變馬爾可夫系統(tǒng)模型引入到序貫多目標(biāo)貝葉斯濾波器中,提出了帶有跳變馬爾可夫系統(tǒng)模型的序貫多目標(biāo)貝葉斯濾波器,并且分別提出了該濾波器在線性高斯系統(tǒng)和非線性高斯系統(tǒng)的實(shí)現(xiàn)方法。仿真實(shí)驗(yàn)結(jié)果表明,在目標(biāo)運(yùn)動模式具有不確性、測量存在不確定性以及存在雜波等情況時,該濾波器能夠?qū)Χ鄼C(jī)動目標(biāo)進(jìn)行有效、穩(wěn)定和準(zhǔn)確的跟蹤。
[Abstract]:The main purpose of multi-target tracking is to measure the target motion pattern with uncertainty, measurement uncertainty and the existence of clutter and so on. It can effectively detect each target and estimate the state of each target. Traditional multi-target tracking methods usually use data association technology. However, data association may have some problems such as "combination explosion" and exponential increase of computational complexity. In order to solve this problem, Mahler proposed a probability assumption density (PHD) based on random finite set. The filter not only avoids data association, but also solves the problem of multi-target tracking in the case of false alarm, missed detection and unknown target number, although PHD filter has many advantages in the process of multi-target tracking. But there are some problems in the filter. Firstly, it is difficult to distinguish the short distance target. Secondly, the filter has a centralized processing of the received measurement data, if the data processing is not timely. Finally, in the case of low detection probability, it is easy to cause the loss of target information and the instability of target number estimation. We propose a sequential multi-target Bayesian filter, which can be used to track multiple maneuvering targets. We propose a sequential multiobjective Bayesian filter for a jump Markov system model. The main content of this paper is as follows: 1) the theory of multiobjective Bayesian filtering based on finite set statistics is introduced. The multi-target tracking model is discussed. The optimal multiobjective Bayesian Bayes is summarized. The filter and the PHD filter with first order moments of joint posteriori distribution are introduced. Finally, the edge distribution of the transfer target and the edge distribution of the existence probability of Bayesian MDBs filter. 2) are introduced. A sequential multiobjective Bayesian filter is proposed. The filter transmits the edge distribution and the probability of existence of the target and processes the measured data received at the current moment sequentially. At the same time. The realization methods of sequential multiobjective Bayesian filter for linear Gao Si system and nonlinear Gao Si system are presented respectively. The simulation results show that when there are clutter, missed detection and unknown number of targets. In order to solve the multi-maneuvering target tracking problem, we introduce the jump Markov system model into sequential multi-target Bayesian filter. A sequential multiobjective Bayesian filter with a jump Markov system model is proposed, and the realization methods of the filter in linear Gao Si system and nonlinear Gao Si system are presented respectively. The simulation results show that the proposed filter can be applied to the system of linear Gao Si and the nonlinear system of Gao Si. When the target motion mode is uncertain, the measurement uncertainty and clutter exist, the filter can track multiple maneuvering targets effectively, stably and accurately.
【學(xué)位授予單位】：深圳大學(xué)
【學(xué)位級別】：碩士
【學(xué)位授予年份】：2017
【分類號】：TN713

【參考文獻(xiàn)】

相關(guān)期刊論文 前6條

1 羅少華;徐暉;薛永宏;;一種新的SMC-PHD濾波的多目標(biāo)狀態(tài)估計(jì)方法[J];宇航學(xué)報(bào);2011年10期

2 吳靜靜;胡士強(qiáng);;基于概率假設(shè)密度的多目標(biāo)視頻跟蹤算法[J];控制與決策;2010年12期

3 歐陽成;姬紅兵;張俊根;;一種改進(jìn)的CPHD多目標(biāo)跟蹤算法[J];電子與信息學(xué)報(bào);2010年09期

4 何佳洲,吳傳利,周志華,陳世福;多假設(shè)跟蹤技術(shù)綜述[J];火力與指揮控制;2004年06期

5 張伯彥,蔡慶宇,袁曾任;一種跟蹤機(jī)動目標(biāo)的新方法[J];系統(tǒng)工程與電子技術(shù);1990年01期

6 周宏仁;機(jī)動目標(biāo)“當(dāng)前”統(tǒng)計(jì)模型與自適應(yīng)跟蹤算法[J];航空學(xué)報(bào);1983年01期

，

本文編號：1392919