Top Banner
WSCG’07 Jonathan Kipling Knight 1 Feb 2007 Copyright © Jonathan Kipling Knight 2007 Fast Skeleton Estimation from Motion Capture Data using Generalized Delogne- Kåsa
25

WSCG’07 Jonathan Kipling Knight 1 Feb 2007 Copyright © Jonathan Kipling Knight 2007 Fast Skeleton Estimation from Motion Capture Data using Generalized.

Dec 18, 2015

Download

Documents

Welcome message from author
This document is posted to help you gain knowledge. Please leave a comment to let me know what you think about it! Share it to your friends and learn new things together.
Transcript
Page 1: WSCG’07 Jonathan Kipling Knight 1 Feb 2007 Copyright © Jonathan Kipling Knight 2007 Fast Skeleton Estimation from Motion Capture Data using Generalized.

WSCG’07

Jonathan Kipling Knight

1 Feb 2007

Copyright © Jonathan Kipling Knight 2007

Fast Skeleton Estimation from Motion Capture Data using Generalized Delogne-Kåsa

Page 2: WSCG’07 Jonathan Kipling Knight 1 Feb 2007 Copyright © Jonathan Kipling Knight 2007 Fast Skeleton Estimation from Motion Capture Data using Generalized.

1 Feb 2007 JKK 2

Synopsis

• Purpose

• Motion Capture

• Skeleton Formation

• Closed Form Solution

• Conclusion

Page 3: WSCG’07 Jonathan Kipling Knight 1 Feb 2007 Copyright © Jonathan Kipling Knight 2007 Fast Skeleton Estimation from Motion Capture Data using Generalized.

1 Feb 2007 JKK 3

Purpose

• Draw an articulated framework of solid segments connected by joints.

• Fastest possible solution from motion capture data

Page 4: WSCG’07 Jonathan Kipling Knight 1 Feb 2007 Copyright © Jonathan Kipling Knight 2007 Fast Skeleton Estimation from Motion Capture Data using Generalized.

1 Feb 2007 JKK 4

Motion Capture

• Magnetic Trackers• Position and Orientation

• Marker Reflectors• Position if in view

• Figure Tracking• Computer vision and image analysis

Page 5: WSCG’07 Jonathan Kipling Knight 1 Feb 2007 Copyright © Jonathan Kipling Knight 2007 Fast Skeleton Estimation from Motion Capture Data using Generalized.

1 Feb 2007 JKK 5

Motion Capture Session

CMU 05-05

QuickTime™ and aYUV420 codec decompressor

are needed to see this picture.

Page 6: WSCG’07 Jonathan Kipling Knight 1 Feb 2007 Copyright © Jonathan Kipling Knight 2007 Fast Skeleton Estimation from Motion Capture Data using Generalized.

1 Feb 2007 JKK 6

Producing a Skeleton

• Single Time Frame• Produce position, size and orientation of

each segment• Markers are fixed 3D positions on segment• Orientation is included with magnetic

trackers• Draw lines between rotation points

Page 7: WSCG’07 Jonathan Kipling Knight 1 Feb 2007 Copyright © Jonathan Kipling Knight 2007 Fast Skeleton Estimation from Motion Capture Data using Generalized.

1 Feb 2007 JKK 7

Time Slice

Page 8: WSCG’07 Jonathan Kipling Knight 1 Feb 2007 Copyright © Jonathan Kipling Knight 2007 Fast Skeleton Estimation from Motion Capture Data using Generalized.

1 Feb 2007 JKK 8

Inverse Kinematics

• What joint angles are needed to get to next position and orientation?

• Good for filling in large frame gaps

• Sometimes more than one answer

Page 9: WSCG’07 Jonathan Kipling Knight 1 Feb 2007 Copyright © Jonathan Kipling Knight 2007 Fast Skeleton Estimation from Motion Capture Data using Generalized.

1 Feb 2007 JKK 9

Inverse Kinematics Example

CMU 05-05

QuickTime™ and a decompressor

are needed to see this picture.

Page 10: WSCG’07 Jonathan Kipling Knight 1 Feb 2007 Copyright © Jonathan Kipling Knight 2007 Fast Skeleton Estimation from Motion Capture Data using Generalized.

1 Feb 2007 JKK 10

Closed Form Solution

• Find centers of rotation for each segment

• Each frame independently drawn

• No iterations

• Quick solution

Page 11: WSCG’07 Jonathan Kipling Knight 1 Feb 2007 Copyright © Jonathan Kipling Knight 2007 Fast Skeleton Estimation from Motion Capture Data using Generalized.

1 Feb 2007 JKK 11

Segment Tree

• Root segment usually hips

• Leaf segments hands, head and feet

• No loops

Root

Leaves

Page 12: WSCG’07 Jonathan Kipling Knight 1 Feb 2007 Copyright © Jonathan Kipling Knight 2007 Fast Skeleton Estimation from Motion Capture Data using Generalized.

1 Feb 2007 JKK 12

Solve Sphere at Each Joint

• One marker on child produces sphere around joint relative to parent

• Must know orientation of parent• 1-3 markers needed or• Magnetic trackers

Page 13: WSCG’07 Jonathan Kipling Knight 1 Feb 2007 Copyright © Jonathan Kipling Knight 2007 Fast Skeleton Estimation from Motion Capture Data using Generalized.

1 Feb 2007 JKK 13

Three Point Orientation

• Three Orthogonal Axes

ˆ z =r p 3 −

r p 1( ) × ˆ x

r p 3 −

r p 1( ) × ˆ x

ˆ x =r p 2 −

r p 1r

p 2 −r p 1

ˆ y = ˆ z × ˆ x

Page 14: WSCG’07 Jonathan Kipling Knight 1 Feb 2007 Copyright © Jonathan Kipling Knight 2007 Fast Skeleton Estimation from Motion Capture Data using Generalized.

1 Feb 2007 JKK 14

Two Point Orientation

• Three Orthogonal Axes

• Substitute center of rotation

ˆ z =r p 2 −

r c ( ) × ˆ x

r p 2 −

r c ( ) × ˆ x

ˆ x =r p 1 −

r c

r p 1 −

r c

ˆ y = ˆ z × ˆ x

Page 15: WSCG’07 Jonathan Kipling Knight 1 Feb 2007 Copyright © Jonathan Kipling Knight 2007 Fast Skeleton Estimation from Motion Capture Data using Generalized.

1 Feb 2007 JKK 15

One Point Orientation

• Three Orthogonal Axes

• Substitute center of rotation and constant axis

ˆ z =r p 1 −

r c ( ) × ˆ x

r p 1 −

r c ( ) × ˆ x

ˆ x = ˆ n

ˆ y = ˆ z × ˆ x

Page 16: WSCG’07 Jonathan Kipling Knight 1 Feb 2007 Copyright © Jonathan Kipling Knight 2007 Fast Skeleton Estimation from Motion Capture Data using Generalized.

1 Feb 2007 JKK 16

Extra Information

• Center of Rotation is available from previously calculated segment

• Constant Axis is available for segments with near cylindrical motion.

Page 17: WSCG’07 Jonathan Kipling Knight 1 Feb 2007 Copyright © Jonathan Kipling Knight 2007 Fast Skeleton Estimation from Motion Capture Data using Generalized.

1 Feb 2007 JKK 17

Center of Sphere

• Generalized Delogne-Kåsa Method for points on a hypersphere xi

rc = x + 1

2C−1S

C = 1N−1 x i − x ( ) x i − x ( )

T

i=1

N

S = 1N−1 x i − x ( ) x i − x ( )

Tx i − x ( )

i=1

N

x = 1N x i

i=1

N

Page 18: WSCG’07 Jonathan Kipling Knight 1 Feb 2007 Copyright © Jonathan Kipling Knight 2007 Fast Skeleton Estimation from Motion Capture Data using Generalized.

1 Feb 2007 JKK 18

Constant Axis

• Test by condition number or determinant of C

• Null vector is axis of cylinder motion

• Center of circle €

C ≈ 0

C ˆ n = 0

r′ c =

r c + ˆ n ̂ n T x −

r c ( )

Page 19: WSCG’07 Jonathan Kipling Knight 1 Feb 2007 Copyright © Jonathan Kipling Knight 2007 Fast Skeleton Estimation from Motion Capture Data using Generalized.

1 Feb 2007 JKK 19

GDK Properties

• Closed form solution for any dimension

• Fastest known 26N

• Cholesky inverse of 3x3 matrix

• Biased when partial coverage of sphere

• As accurate as data O()

Page 20: WSCG’07 Jonathan Kipling Knight 1 Feb 2007 Copyright © Jonathan Kipling Knight 2007 Fast Skeleton Estimation from Motion Capture Data using Generalized.

1 Feb 2007 JKK 20

Marker Requirements

• 3 Markers on root segment of tree

• 1-3 Markers on all other segments

• Segments with 1 Marker should have one degree of freedom (e.g.knee,elbow)

Page 21: WSCG’07 Jonathan Kipling Knight 1 Feb 2007 Copyright © Jonathan Kipling Knight 2007 Fast Skeleton Estimation from Motion Capture Data using Generalized.

1 Feb 2007 JKK 21

Break Dance

CMU 85-14

QuickTime™ and a decompressor

are needed to see this picture.

Page 22: WSCG’07 Jonathan Kipling Knight 1 Feb 2007 Copyright © Jonathan Kipling Knight 2007 Fast Skeleton Estimation from Motion Capture Data using Generalized.

1 Feb 2007 JKK 22

Salsa Dance

CMU 60-08

QuickTime™ and a decompressor

are needed to see this picture.

Page 23: WSCG’07 Jonathan Kipling Knight 1 Feb 2007 Copyright © Jonathan Kipling Knight 2007 Fast Skeleton Estimation from Motion Capture Data using Generalized.

1 Feb 2007 JKK 23

Conclusion

• GDK is fastest available sphere solution 26N

• As accurate as data O()

• 1-3 Marker requirements per segment

• Provides skeleton to attach solid shape

Page 24: WSCG’07 Jonathan Kipling Knight 1 Feb 2007 Copyright © Jonathan Kipling Knight 2007 Fast Skeleton Estimation from Motion Capture Data using Generalized.

1 Feb 2007 JKK 24

Future Research

• Unbiased version of GDK

• Full analysis of statistical nature

• Condition for acceptable data

Page 25: WSCG’07 Jonathan Kipling Knight 1 Feb 2007 Copyright © Jonathan Kipling Knight 2007 Fast Skeleton Estimation from Motion Capture Data using Generalized.

1 Feb 2007 JKK 25

Acknowledgments

• The data used in this project was obtained from mocaps.cs.cmu.edu. The database was created with funding from NSF EIA-0196217.