Sequential Monte-Carlo Method -Introduction, implementation and application Fan, Xin 2005.3.28.

Slides:



Advertisements
Similar presentations
Discussion for SAMSI Tracking session, 8 th September 2008 Simon Godsill Signal Processing and Communications Lab. University of Cambridge www-sigproc.eng.cam.ac.uk/~sjg.
Advertisements

Bayesian Belief Propagation
Motivating Markov Chain Monte Carlo for Multiple Target Tracking
State Estimation and Kalman Filtering CS B659 Spring 2013 Kris Hauser.
CSCE643: Computer Vision Bayesian Tracking & Particle Filtering Jinxiang Chai Some slides from Stephen Roth.
Markov Localization & Bayes Filtering 1 with Kalman Filters Discrete Filters Particle Filters Slides adapted from Thrun et al., Probabilistic Robotics.
Introduction to Sampling based inference and MCMC Ata Kaban School of Computer Science The University of Birmingham.
Oklahoma State University Generative Graphical Models for Maneuvering Object Tracking and Dynamics Analysis Xin Fan and Guoliang Fan Visual Computing and.
PHD Approach for Multi-target Tracking
Probabilistic Robotics Bayes Filter Implementations Particle filters.
Particle Filters.
Particle Filters Pieter Abbeel UC Berkeley EECS Many slides adapted from Thrun, Burgard and Fox, Probabilistic Robotics TexPoint fonts used in EMF. Read.
Stanford CS223B Computer Vision, Winter 2007 Lecture 12 Tracking Motion Professors Sebastian Thrun and Jana Košecká CAs: Vaibhav Vaish and David Stavens.
CS 547: Sensing and Planning in Robotics Gaurav S. Sukhatme Computer Science Robotic Embedded Systems Laboratory University of Southern California
Sérgio Pequito Phd Student
Nonlinear and Non-Gaussian Estimation with A Focus on Particle Filters Prasanth Jeevan Mary Knox May 12, 2006.
Particle filters (continued…). Recall Particle filters –Track state sequence x i given the measurements ( y 0, y 1, …., y i ) –Non-linear dynamics –Non-linear.
Object Detection and Tracking Mike Knowles 11 th January 2005
Particle Filters for Mobile Robot Localization 11/24/2006 Aliakbar Gorji Roborics Instructor: Dr. Shiri Amirkabir University of Technology.
Probabilistic Robotics Introduction Probabilities Bayes rule Bayes filters.
1 Integration of Background Modeling and Object Tracking Yu-Ting Chen, Chu-Song Chen, Yi-Ping Hung IEEE ICME, 2006.
Today Introduction to MCMC Particle filters and MCMC
Comparative survey on non linear filtering methods : the quantization and the particle filtering approaches Afef SELLAMI Chang Young Kim.
Probabilistic Robotics Bayes Filter Implementations Particle filters.
Stanford CS223B Computer Vision, Winter 2007 Lecture 12 Tracking Motion Professors Sebastian Thrun and Jana Košecká CAs: Vaibhav Vaish and David Stavens.
Particle Filtering for Non- Linear/Non-Gaussian System Bohyung Han
© 2003 by Davi GeigerComputer Vision November 2003 L1.1 Tracking We are given a contour   with coordinates   ={x 1, x 2, …, x N } at the initial frame.
Sampling Methods for Estimation: An Introduction
Novel approach to nonlinear/non- Gaussian Bayesian state estimation N.J Gordon, D.J. Salmond and A.F.M. Smith Presenter: Tri Tran
Bayesian Filtering for Location Estimation D. Fox, J. Hightower, L. Liao, D. Schulz, and G. Borriello Presented by: Honggang Zhang.
Real-Time Decentralized Articulated Motion Analysis and Object Tracking From Videos Wei Qu, Member, IEEE, and Dan Schonfeld, Senior Member, IEEE.
Particle Filtering. Sensors and Uncertainty Real world sensors are noisy and suffer from missing data (e.g., occlusions, GPS blackouts) Use sensor models.
HCI / CprE / ComS 575: Computational Perception
Bayesian Filtering for Robot Localization
Particle Filtering in Network Tomography
Markov Localization & Bayes Filtering
BraMBLe: The Bayesian Multiple-BLob Tracker By Michael Isard and John MacCormick Presented by Kristin Branson CSE 252C, Fall 2003.
Object Tracking using Particle Filter
Computer vision: models, learning and inference Chapter 19 Temporal models.
From Bayesian Filtering to Particle Filters Dieter Fox University of Washington Joint work with W. Burgard, F. Dellaert, C. Kwok, S. Thrun.
SIS Sequential Importance Sampling Advanced Methods In Simulation Winter 2009 Presented by: Chen Bukay, Ella Pemov, Amit Dvash.
Particle Filtering (Sequential Monte Carlo)
Computer vision: models, learning and inference Chapter 19 Temporal models.
Jamal Saboune - CRV10 Tutorial Day 1 Bayesian state estimation and application to tracking Jamal Saboune VIVA Lab - SITE - University.
Probabilistic Robotics Bayes Filter Implementations.
Overview Particle filtering is a sequential Monte Carlo methodology in which the relevant probability distributions are iteratively estimated using the.
Particle Filters.
Mixture Models, Monte Carlo, Bayesian Updating and Dynamic Models Mike West Computing Science and Statistics, Vol. 24, pp , 1993.
Forward-Scan Sonar Tomographic Reconstruction PHD Filter Multiple Target Tracking Bayesian Multiple Target Tracking in Forward Scan Sonar.
Mobile Robot Localization (ch. 7)
Processing Sequential Sensor Data The “John Krumm perspective” Thomas Plötz November 29 th, 2011.
Maximum a posteriori sequence estimation using Monte Carlo particle filters S. J. Godsill, A. Doucet, and M. West Annals of the Institute of Statistical.
Michael Isard and Andrew Blake, IJCV 1998 Presented by Wen Li Department of Computer Science & Engineering Texas A&M University.
An Introduction to Kalman Filtering by Arthur Pece
Mixture Kalman Filters by Rong Chen & Jun Liu Presented by Yusong Miao Dec. 10, 2003.
CSE-473 Project 2 Monte Carlo Localization. Localization as state estimation.
Short Introduction to Particle Filtering by Arthur Pece [ follows my Introduction to Kalman filtering ]
OBJECT TRACKING USING PARTICLE FILTERS. Table of Contents Tracking Tracking Tracking as a probabilistic inference problem Tracking as a probabilistic.
Tracking with dynamics
Introduction to Sampling Methods Qi Zhao Oct.27,2004.
The Unscented Particle Filter 2000/09/29 이 시은. Introduction Filtering –estimate the states(parameters or hidden variable) as a set of observations becomes.
CS Statistical Machine learning Lecture 25 Yuan (Alan) Qi Purdue CS Nov
Rao-Blackwellised Particle Filtering for Dynamic Bayesian Networks Arnaud Doucet, Nando de Freitas, Kevin Murphy and Stuart Russell CS497EA presentation.
Particle filters for Robot Localization An implementation of Bayes Filtering Markov Localization.
Probabilistic Robotics Bayes Filter Implementations Gaussian filters.
Particle Filtering for Geometric Active Contours
Probabilistic Robotics
Course: Autonomous Machine Learning
Dynamical Statistical Shape Priors for Level Set Based Tracking
Filtering and State Estimation: Basic Concepts
Presentation transcript:

Sequential Monte-Carlo Method -Introduction, implementation and application Fan, Xin

Probabilistic state estimation for a dynamic system Dynamic system, a system with changes over time What can SMC do -Economics, weather -Moving object, image -Generally speaking, anything in the world Extracting relevant information of the system through investigating the observations

State, hidden information to describe the system What can SMC do--State-Space Modeling -Kinematic characteristics in tracking Measurements, made on the system —observed noisy data -Image data available up to current time —Evolving over time (Dynamic model): —What we are interested -Intensities of pixels in image estimation -Intensities of the degraded image —Associated with states (Measurement Model):

State evolution is described in terms of transition probability What can SMC do--Probabilistic formulation How the given fits the available measurement is described in terms of likelihood probability Determining the belief in the state taking deferent values, given the measurements

Prediction: What can SMC do—Recursive Estimation Update with the innovative measurement Starting from, at time is estimated with available :

Why use SMC Only when all of the distributions are Gaussian, the posterior distribution is Gaussian and analytical solution exists -- Kalman filter Non-Gaussian process noise Nonlinear Dynamics --sudden and jerky motion Multiple targets tracking Partial occlusion

Implementation—Basic idea Use SAMPLES with associated weights to approximate posterior density Examples: --discrete probability: coin, galloping dominoes --continuous density sampling Gaussion density

Implementation—Basic assumptions No explicit assumptions on the forms of both transition and likelihood probabilities, SMC is applicable for nonlinear and non-Gaussian estimation Measurements are independent, both mutually and with respect to dynamical process: Markov Chain:

Implementation—A 1D nonlinear example We need to infer the state at the time with available measurements Measurements Model: Dynamic Model:

Implementation—Results

Implementation—Algorithm Initialization -Draw samples from -Set weights 1. Prediction: 2. Update: -Normalization - 3. Resample

Implementation—Results

Implementation—Discussion Relaxes : - Linearity of dynamic and measurement models - The forms of the distributions of process and measurement noise. Requires : - Initial prior density - The likelihood can be evaluated - State samples can be generated easily  Do not make use of any knowledge of the measurements  inefficient and sensitive to outliers

Implementation—Generic SIS Algorithm - Draw 1. Prediction: 2. Update: -Normalization - 3. Resample Introducing an Importance density to facilitate sampling and using observations

Application—Contour extraction Probabilistic state estimation formulation: Problem Definition: - Grouping edge points into continuous cures, represented by a series of control points. - The positions of the control points are the states, then a contour turns out to be a state sequence. - Edge points are those pixels with larger intensity gradients, which are used as measurements

Application—Contour extraction Definitions of the probabilities Likelihood: Dynamics: Importance density: Perform the standard procedure to estimate the states

Application—Some results

Summary of using SMC Define the probability densities Modeling problems as probabilistic estimation -States / what we want, but cannot observe directly -Measurements / observations - Likelihood / the relationship between states and measurements / functional form that can be evaluated - Transition / determine the evolution of the states over time / the prior knowledge of the system under investigation - Importance / employ the observations / easy for sampling

Future work Apply SMC to various problems - Vision tracking - Constrain the state space by using better dynamic model / incorporate more prior knowledge - Elaborate techniques for efficiently sampling / SA / move samples to density peaks - Data fusion - Image restoration/super-resolution - Digital communication High computational expense - Decompose a high dimensional problem to several lower dimensional ones…

Reference [4] P. Pérez, A. Blake, and M. Gangnet. JetStream: Probabilistic contour extraction with particles. Proc. Int. Conf. on Computer Vision (ICCV), II: , Contour extraction [3] Gordon, N., Salmond, D., and Smith, A.." Novel approach to nonlinear/non-Gaussian Bayesian state estimation". IEE Proc. F, 140, 2, the simple 1D example [1] Proceedings of the IEEE, vol. 92, no. 3, Mar Special issue [2] IEEE Trans On Signal Processing, Vol. 50, no. 2. Special issue [5] M. Isard and A. Blake, "Contour tracking by stochastic propagation of conditional density", ECCV96,pp ,1996. – Application to vision tracking, in which significant performance was achieved. [6] Jun S. Liu and Rong Chen, "Sequential Monte Carlo Methods for Dynamic Systems", Journal of the American Statistical Association, Vol. 93, No. 443, pp , – SMC from the point of statisticians