35
Real-time 3D Object Pose Estimation and Tracking for Natural Landmark Based Visual Servo Seung-Min Baek and Sukhan Lee Sungkyunkwan University Intelligent System Research Center Changhyun Choi Georgia Tech College of Computing

IEEE/RSJ IROS 2008 Real-time Tracker

  • Upload
    cchoi

  • View
    1.276

  • Download
    0

Embed Size (px)

DESCRIPTION

This slides were for presentation in IROS 2008 conference.

Citation preview

Page 1: IEEE/RSJ IROS 2008 Real-time Tracker

Real-time 3D Object Pose Estimation and Tracking for Natural Landmark Based Visual Servo

Seung-Min Baek and Sukhan LeeSungkyunkwan University

Intelligent System Research Center

Changhyun ChoiGeorgia Tech

College of Computing

Page 2: IEEE/RSJ IROS 2008 Real-time Tracker

Contents• Introduction– Motivation– Related Works

• Proposed Approach– System Overview– Problem Definition– Initial Pose Estimation– Local Pose Estimation

• Experimental Results• Summary & Conclusion• Future Work

IEEE/RSJ IROS 2008, Sept 25 2

Page 3: IEEE/RSJ IROS 2008 Real-time Tracker

Introduction

In Visual Servo Control,• Object Recognition • Pose Estimation are key tasks.

IEEE/RSJ IROS 2008, Sept 25 3

Page 4: IEEE/RSJ IROS 2008 Real-time Tracker

Introduction

Many systems still useArtificial Landmark.

Unnatural in human environment

IEEE/RSJ IROS 2008, Sept 25 4

Page 5: IEEE/RSJ IROS 2008 Real-time Tracker

Introduction

We need Natural Landmarks.

Natural Landmarks are visual features objects inherently have.IEEE/RSJ IROS 2008, Sept 25 5

Page 6: IEEE/RSJ IROS 2008 Real-time Tracker

Introduction

Modern recognition methods

SIFTabout 200~300 ms on a modern PC

Structured lightseveral seconds

IEEE/RSJ IROS 2008, Sept 25 6

Page 7: IEEE/RSJ IROS 2008 Real-time Tracker

Motivation

• How to apply these state-of-the-art recognition methods to visual servo control?

• How to overcome the time lag?

• How to solve the real-time issue?

IEEE/RSJ IROS 2008, Sept 25 7

Page 8: IEEE/RSJ IROS 2008 Real-time Tracker

Related Works

• Monocular • Model-based• Use keyframe information

as prior knowledge• Use sparse bundle

adjustment technique[ L. Vacchetti et al., PAMI 04 ]

Input image should be close enough to the prior knowledge!

IEEE/RSJ IROS 2008, Sept 25 8

Page 9: IEEE/RSJ IROS 2008 Real-time Tracker

Related Works

• Active Contour• Local curve fitting

algorithm• Initialize by SIFT keypoint

matching

[G. Panin and A. Knoll, JMM 04 ]

Potential danger in background having same color with tracking object!

IEEE/RSJ IROS 2008, Sept 25 9

Page 10: IEEE/RSJ IROS 2008 Real-time Tracker

Our Idea

• Use prior knowledge (object models)– 2D images– 3D points obtained from structured light system

• Use scale invariant feature matching for accurate initialization

• Use KLT (Kanade-Lucas-Tomasi) tracker for fast local tracking

IEEE/RSJ IROS 2008, Sept 25 10

Page 11: IEEE/RSJ IROS 2008 Real-time Tracker

System Overview

• Add text

IEEE/RSJ IROS 2008, Sept 25 11

Page 12: IEEE/RSJ IROS 2008 Real-time Tracker

Two Modes

• Mono Mode– Using mono camera– Better computational performance

• Stereo Mode– Using stereo camera– More accurate pose result

IEEE/RSJ IROS 2008, Sept 25 12

Page 13: IEEE/RSJ IROS 2008 Real-time Tracker

Problem Definition – Mono Mode

Given 2D-3D correspondences and a calibrated mono camera, find the pose of the object with

respect to the camera.IEEE/RSJ IROS 2008, Sept 25 13

Page 14: IEEE/RSJ IROS 2008 Real-time Tracker

Problem Definition – Stereo Mode

Given 3D-3D correspondences and a calibrated stereo camera, find the pose of the object

with respect to the camera.IEEE/RSJ IROS 2008, Sept 25 14

Page 15: IEEE/RSJ IROS 2008 Real-time Tracker

Initial Pose Estimation

• Add text

IEEE/RSJ IROS 2008, Sept 25 15

Page 16: IEEE/RSJ IROS 2008 Real-time Tracker

Initial Pose Estimation

1. Extract SIFT keypoints2. Matching with model

knowledge3. Estimate initial pose4. Get a convex hull of a set of

matched SIFT keypoints5. Generate KLT tracking points

within the convexhull6. Calculate 3D coordinates of

KLT points

IEEE/RSJ IROS 2008, Sept 25 16

Page 17: IEEE/RSJ IROS 2008 Real-time Tracker

Initial Pose Estimation

• Mono Mode– Use the POSIT algorithm (2D-

3D)

• Stereo Mode– Use the closed-form solution

using unit quaternions (3D-3D)

R,tR,t

R,tR,t

IEEE/RSJ IROS 2008, Sept 25 17

Page 18: IEEE/RSJ IROS 2008 Real-time Tracker

Initial Pose Estimation

1. Extract SIFT keypoints2. Matching with model

knowledge3. Estimate initial pose4. Get a convex hull of a set of

matched SIFT keypoints5. Generate KLT tracking points

within the convexhull6. Calculate 3D coordinates of

KLT points

IEEE/RSJ IROS 2008, Sept 25 18

Page 19: IEEE/RSJ IROS 2008 Real-time Tracker

Initial Pose Estimation• 3D coordinates of each KLT points are required for

subsequent local pose estimation

• Stereo Mode– Straightforward in a calibrated stereo rig– Triangulate 3D points

• Mono Mode– Use approximation with the knowledge of model– Get 3D coordinates by using three nearest neighboring

SIFT points

IEEE/RSJ IROS 2008, Sept 25 19

Page 20: IEEE/RSJ IROS 2008 Real-time Tracker

Initial Pose Estimation

+ : SIFT points• : KLT points

IEEE/RSJ IROS 2008, Sept 25 20

Page 21: IEEE/RSJ IROS 2008 Real-time Tracker

Initial Pose Estimation

Treat the surface as locally flatIEEE/RSJ IROS 2008, Sept 25 21

Page 22: IEEE/RSJ IROS 2008 Real-time Tracker

Local Pose Estimation

• Add text

IEEE/RSJ IROS 2008, Sept 25 22

Page 23: IEEE/RSJ IROS 2008 Real-time Tracker

Local Pose Estimation

• Estimate pose with KLT tracking points and their 3D points

• Pose estimation algorithms are same– Mono Mode

• Use the POSIT algorithm (2D-3D)

– Stereo Mode• Use the closed-form solution using unit

quaternions (3D-3D)

R,tR,t

R,tR,t

IEEE/RSJ IROS 2008, Sept 25 23

Page 24: IEEE/RSJ IROS 2008 Real-time Tracker

Removing Outliers

IEEE/RSJ IROS 2008, Sept 25 24

Page 25: IEEE/RSJ IROS 2008 Real-time Tracker

Outlier Handling• KLT tracking points are easy to

drift

• Drifting points result in inaccurate pose

• Use RANSAC to remove outlier

• Re-initialize when there are no sufficient # of inliers

IEEE/RSJ IROS 2008, Sept 25 25

Page 26: IEEE/RSJ IROS 2008 Real-time Tracker

Tracking Results

IEEE/RSJ IROS 2008, Sept 25 26

Page 27: IEEE/RSJ IROS 2008 Real-time Tracker

Experiment

Mono Mode

Stereo ModeIEEE/RSJ IROS 2008, Sept 25 27

Page 28: IEEE/RSJ IROS 2008 Real-time Tracker

Tracking Results - translation

IEEE/RSJ IROS 2008, Sept 25 28

Page 29: IEEE/RSJ IROS 2008 Real-time Tracker

Tracking Results - rotation

IEEE/RSJ IROS 2008, Sept 25 29

Page 30: IEEE/RSJ IROS 2008 Real-time Tracker

RMS Error

RMS errors over the whole sequence of image

Z

IEEE/RSJ IROS 2008, Sept 25 30

Page 31: IEEE/RSJ IROS 2008 Real-time Tracker

Computational Time

Computational times of pose estimationIEEE/RSJ IROS 2008, Sept 25 31

Page 32: IEEE/RSJ IROS 2008 Real-time Tracker

Computational Time

Computational times of each module

IEEE/RSJ IROS 2008, Sept 25 32

Page 33: IEEE/RSJ IROS 2008 Real-time Tracker

Summary & Conclusion

• A method for tracking 3D roto-translation of rigid objects – using scale invariant feature based matching – KLT (Kanade-Lucas-Tomasi) tracker

• Mono mode– guarantees higher frame rate performance

• stereo mode– shows better pose results

IEEE/RSJ IROS 2008, Sept 25 33

Page 34: IEEE/RSJ IROS 2008 Real-time Tracker

Future Work

• To decrease the computational burden– Use GPU-based implementation of KLT tracker and

SIFT• GPU KLT• SiftGPU

– Unifying the contour based tracking

IEEE/RSJ IROS 2008, Sept 25 34

Page 35: IEEE/RSJ IROS 2008 Real-time Tracker

Thank you

• Any Questions?

• Any Suggestions?

• Any Comments?

IEEE/RSJ IROS 2008, Sept 25 35