优化无味粒子滤波器：电经纬仪视觉追踪的高效解决方案

174 浏览量更新于2024-08-26 收藏 236KB PDF 举报

本文探讨了在电经纬仪的视觉追踪系统中应用无味粒子过滤器（Unscented Particle Filter, UPF）的适配技术，以满足高精度跟踪需求的同时降低计算复杂性。电经纬仪是一种精密测量仪器，其追踪过程通常涉及线性状态空间模型，这对于采用粒子滤波算法非常有利。作者们首先注意到，无味粒子滤波器的核心在于其 Unscented Transform（无迹变分法），这是一种处理非线性问题的有效手段。在电经纬仪的视觉追踪中，通过简化这一过程，他们将Sigmasampling（扩散采样）分解为一个简单的组合步骤，这显著减少了计算步骤，提高了效率。进一步地，他们提出了一种混合系统UPF（Mixed System Unscented Particle Filter）算法，对非线性信号采样、状态向量、测量向量以及它们的协方差矩阵进行处理，这种创新的方法使得在保持跟踪精度的同时，计算负担得以大幅度减轻。与传统的UPF相比，实验结果显示，提出的算法在计算准确性上有显著提升，同时计算工作量明显减少。总结来说，本研究针对电经纬仪视觉追踪的具体应用场景，巧妙地融合了无味粒子过滤器的理论和技术，通过简化操作流程和引入混合系统UPF，成功实现了高效且精确的追踪性能。这对于提高电经纬仪在导航、测量等领域的实际应用性能具有重要的意义，也为其他领域中的视觉跟踪系统提供了新的优化思路。

Adaption of Unscen

ted Particle filter for Visual

tracking in Electro-Optic Theodolite

ongtai Cai, Rui Zhang ,Qingxiang Wu

Coll

ege of Photonic and Electronic Engineering,

Fujian Normal University

Fuzhou, 350108, China

{gjrtcai, qxwu}@163.com

Honghai Sun

Changchu

n Institute of Optics, Fine Mechanics and

Physics,

Chinese Academy of Science

Changchun, 130033, China

{HHSun }@ciomp.ac.cn

Abstra

ct—To satisfy the need of visual tracking in electro-optic

theodolite, we use some simplification technology to adapt UPF

(Unscented Particle Filter), which reduce the computational

complexity considerably. The state space equation in electro-

optic theodolite tracking system is linear; the sigma sampling

in unscented transform can be simplified as a decomposition-

composition process; the nonlinear transmit of sigma sampling,

state vector, measurement vector and their covariance matrix

are simplified by an MSUPF (UPF for Mixture system)

algorithm. Experiment result shows that the proposed

algorithm has better calculation accuracy with considerable

lower computational complexity compare with UPF. The

computation work of the proposed algorithm is only 81.13% of

that of the UPF.

Keywords- particle filter; unscented particle filter; visual

tracking; object tracking

TRODUCTION

we know, the academic society and the industry

society are considering different aspect of an algorithm. So

the adaptation of an algorithm derives from academic society

to satisfy the need of industry application is an important

topic

[

Bay

esian filter based algorithm is one kind of important

methods in visual tracking, especially the Particle filters (PFs)

[2]

which have

large spectrum of applications in visual

tracking. But the PFs in academic society are not cut into the

industry shape. So we adapt the traditional PFs in academy

society to satisfy the need of the industry application.

II. M

OTIO

ODEL

CKED

BJECT

simple visual tracking applications, the location of

tracked object is recorded with X-axis and Y-axis in

orthogonal coordinates, but the location of tracked object in

electro-optic theodolite system is recorded with angle of

orientation, angle of pitch and radial distance in spherical

coordinates.

Suppose that the tracked object is moving in a smooth

trajectory, the motion trace can be described with a

polynomial curve. Specifically, a second order polynomial is

employed to describe the motion trace of the tracked object.

In orthogonal coordinates, that is

0 1 2

( ) ( )

x t a a t a t n t

   

(1)

0 1 2

( ) ( )

y t b b t b t n t

   

(2)

0 1 2

( ) ( )

z t c c t c t n t

   

(3)

where x(t), y(t) and z(t) is the position of the object in x-axis,

y-axis, and z-axis respectively; n

(t),

(t)

and n

(t) are

noise

respectively

[

Acc

ording to (1), (2) and (3), the state equation of the

tracked object is

( 1) ( ) ( )

k k k

  x Ax n

(4)

where

 

( ) ( ) ( ) ( ) ( ) ( ) ( ) ( ) ( ) ( )

k x k x k x k y k y k y k z k z k z k

x

   

 

is th

e state parameters,

, , , , , , , ,x x x y y y z z z

   

 

are

position,

speed and acceleration components in x-axis, y-axis and z-

axis respectively.

is state transition matrix:

2 2 2

1 1 1

2 2 2

1 1 1

0 1 0 1 0 1

0 0 1 0 0 1 0 0 1

T T T T T T

T T T

     

     

  

     

     

, T

is the

sampling period, And suppose that

(k)

is Gaussian white

noise with zero-mean. k is time variable. The correlation

matrix of the noise is defined by

 

( ) 1 1 1

nx xx

k diag



 R



where



is corr

elation coefficient of noise.

According to the relationship of spherical coordinates

and rectangular coordinates, the measurement equation is

2 2

2 2 2

( )

( ) ( )

( ) ( ) ( ) ( ) ( )

( ) arctan ( )

( ) ( )

arctan

y k

x k

z k

x k y k

R k x k y k z k n k

V k n k

H k n k



 

 

   

 

   

 

 

   

 

   

 

   

 

, (5)

where x(k), y(k)and z(k) are position components in x-axis,

y-axis and z-axis respectively, R(k),V(k) and H(k) are radial

distance, azimuth angle and elevation angle respectively,

(t), n

(t) and n

(t) are noise in radial distance, azimuth

angle and elevation angle respectively.

For conciseness, (5) can be written as

( ) [ ( ), ] ( )k h k k k

 z x n

(6)

181

ICWMMN2013 Proceedings

下载后可阅读完整内容，剩余3页未读，立即下载

weixin_38723527

粉丝: 3
资源: 953

优化无味粒子滤波器：电经纬仪视觉追踪的高效解决方案

unscentedParticleFilter:根据 Van Der Merwe 等人的论文实施无味粒子过滤器

分布式电驱动车辆的无味粒子滤波状态参数联合观测.pdf

pyfill:pyfill是用于传感器融合算法的静态python软件包，除运动模型和测量模型外，它还包含最常见的概率过滤器（例如线性卡尔曼，扩展卡尔曼，无味卡尔曼，粒子过滤器），易于使用且经过全面测试

SFND_Unscented_Kalman_Filter：用于无度纳米级项目的无味卡尔曼过滤器

非高斯条件下基于无味粒子滤波器的目标跟踪 (2007年)

基于无味强跟踪信息过滤器的多传感器融合

基于交互多模型的机动目标跟踪的无味信息共识过滤器

基于无味粒子滤波和交互多模型算法的多机动目标跟踪 (2008年)

数据融合matlab代码-CarND-Unscented-Kalman-Filter:无味卡尔曼过滤器项目

无味：灵活且功能强大的无味卡尔曼过滤器C ++ 11库，不对您所估计的内容或测量方式进行任何假设

最新资源