MIMIC: REAL-TIME MARKER-FREE MOTION

CAPTURE SYSTEM TO CREATE AN AGENT IN THE

VIRTUAL SPACE

Team members

 Rashmi M S
 Vidya Kori

 Basavaraj B N

 Shivamurthi
ABSTRACT
A real-time 3D computer vision system called
MIMIC that can capture and save motion of an actor.
This system analyzes input images from vision
sensors and searches feature information like a head,
hands, and feet. this estimates intermediated joints as an
elbow and knee using feature information and makes 3D
human model having 20 joints. This virtual human
model mimics the motion of an actor in realtime.
INTRODUCTION
 Capture movements of human in advance and mapping
those into characters. This is generally known as
“Motion capture” method.
 After marking to an actor, we capture movements of
markers using infrared cameras around the actor.
 Analyzing movements of marker, we obtain information
of the movements of each joint of an actor and applying
this information to the virtual character, obtain the
results of natural movement like a real actor.
MIMIC-SYSTEM CONFIGURATION
 The MIMIC consists of one server and two clients.
 Each client has one CCD camera, and transmits the
results of feature extraction to the server after analyzing
inputted image.
 Cameras are installed front-left and front-right of an
actor.
 The server performs reconstruct 3D features from 2D
features information, track the 3D features, calculate the
intermediated joints, and show the result of human
model.
ESTIMATION OF 2D POSITION OF
FEATURES
 Acquire actor’s images using synchronized two color
CCD cameras.
 Extract a color range, a region property of a body and an
end-effector of human by segmenting a difference image
of a background and an initial image into color region.
 After sorting substitute pixels into a color range in
successive image of movements , extract the ratio of
pixels which are existed a circumscribed rectangle of a
feature region, a perimeter, a size of a feature region
through a fast run-length labeling algorithm.
ESTIMATION OF 3D POSITION OF
FEATURES
 As body features have unrestrained movement, however,
have to track the movement of each features in order to
match those features correctly. For this work, we use the
Kalman filter.
 Estimate next position value from 3D position of
obtained features in the current time.