Motion Capture. Motion Capture in Movies. Motion Capture in Games

Transcription

1 Motion Capture Motion Capture in Movies 2 Motion Capture in Games 3 4

2 Magnetic Capture Systems Tethered Sensitive to metal Low frequency (60Hz) Mechanical Capture Systems Any environment Measures joint angles Restricts the motion 5 6 Optical motion capture Place markers on the actor Optical Capture Systems 8 or more cameras Restricted volume High Frequency (240Hz) Occlusions Cameras can determine marker positions 7 8

3 How Does It Work? Optical motion capture process 8 cameras Hz + Special tape = Raw Point Data 3. Compute marker positions from camera images 9 10 Optical motion capture process Problem Statement 3. Compute marker positions from camera images 11 12

4 Automatic Calibration Design Goals: Fully automatic Any skeleton Accurate Generic Skeleton Actor s kinematics structure, and rough handle positions Input Calibration Data Initial path data that exercises all of the subject s DOFs Independent Variables DOFs Bone lengths Handle offsets Global scale Optical motion capture process 3. Compute marker positions from camera images 15 16

5 Optical motion capture process Marker Identification 3. Compute marker paths from camera images (X 0, Y 0, Z 0 ) (X 1, Y 1, Z 1 ) (X 2, Y 2, Z 2 ) At each frame, motion capture gives us a set of points θ We would like something more intuitive Marker Identification Problems Optical motion capture process 3. Compute marker positions from camera images Making sense of raw data 19 20

6 IK Problem Definition Inverse Kinematics 1. Create a handle on body position or orientation 2. Pull on the handle 3. IK figures out how joint angles should change Inputs: An articulated skeleton with handles. Desired positions for handles. Outputs: Joint angles that move handles to desired positions Inverse Kinematics (con t( con t) We are solving IK on a complex model (~50 DOFs and 30 handles). Motion capture data often contains missing markers. Many different formulations for IK problem, would like to use one that is best for motion capture data. Let: Then: q More Formally actor state vector (joint bundle) C(q) constraint functions that pull handles solve for q such that C(q) =

7 What s a Constraint? Constraint derivatives q=[ x h,θ h, φ h, σ h, θ t, φ t, σ t, θ c, θ f, φ f ] x h, θ h, φ h, σ h θ t, φ t, σ t θ c θ f, φ f desired position d handle h(q) Can be rich, complicated But most common is very simple: Position constraint just sets difference of two vectors to zero: C(q) = h(q) - d = 0 q=[ x h, t, c, f, f ] x h t c f, f desired position d handle h(q) C(q) = h(q) --d = 0 C( q) h( q) = q q Computing Derivatives Jacobian Matrix h θ x h t w c c f, f hv s z y Apply the chain rule Need to know how to compute derivatives for each transformation h s x h = T( x, y, z ) R( θ, φ, σ ) TR( θ, φ, σ ) TR( θ ) TR( θ, φ ) h w h h h h h h t t t c f f R( θc ) = T( x, y, z ) R( θ, φ, σ ) TR( θ, φ, σ ) T TR( θ, φ ) h h h h h h h t t t f f θ c s 27 s C q e θ e x. 0. y. 1. z. 0. C e handle C Can compute Jacobian for each constraint / handle Value of Jacobian depends on current state Jacobian linearly relates joint angle velocity to constraint velocity 28

8 Unconstrained Optimization Real-time Motion Capture 2 Minimize G ( q) G( q) wc i i( q) i Move in the direction of the objective function gradient: G G Ci = + 2 wc i i = + q q i q G q= qo + α q Calibration Generic Skeleton Sample Motion Match markers to handles. Calibrate the skeleton. Begin capture. Initialize marker match. Compute joint angles with IK. Update marker match. Real-time Motion capture as UI Use acting for animation interface Motion Transformation Start with a mocap sequence Edit it to fit the needs of the animation Try to be as close to the original motion as possible 31 32