[其它] A Fully Automatic Rigging Algorithm for 3D Character Animation

[复制链接]

彬彬

797 主题	1 听众	1万积分

资深设计师

Rank: 7 Rank: 7 Rank: 7

纳金币: 5568
精华: 0

电梯直达

楼主

发表于 2011-12-30 18:56:06 |只看该作者 |倒序浏览

Abstract

This paper proposes an automatic algorithm to generate an inverse

kinematic (IK) skeleton for 3D characters. First, a curve skeleton is

extracted from the volume inside the character mesh. The extracted

curve skeleton is then analyzed and classified by comparing it with

the characteristics of templates for real animals. The outcome of

this classification step is an associated class for the given 3D character

model, together with the meaning of each skeleton segment.

Next, each bone of the template skeleton is located in the appropriate

skeleton segment of the input skeleton graph. Unlike previous

methods, this algorithm does not require an original 3D character

model that is created with a restricted pose, topology and orientation.

Furthermore, all stages of the process are completed without

user intervention.

CR Categories: I.3.7 [Computer Graphics]: Three-Dimensional

Graphics and Realism—Animation;

Keywords: automatic rigging, skeleton, animation, template

1 Introduction

Rigging is a laborious process, usually involving skilled users. Automatic

rigging algorithms are being developed to overcome this

problem. Template-based automatic rigging algorithm[Aujay et al.

2007; Baran and Popovi´c 2007] is one of several approaches which

are used to automatically generate an IK skeleton. A predefined

template skeleton is located on the extracted curve skeleton to create

an IK skeleton. Several items of information about each joint

of the template skeleton can then be transferred to the created IK

skeleton via this correspondence. Although previous works have

described the correct generation of an IK skeleton, user intervention

is still required.

Our proposed method also uses the strategy of a template-based algorithm.

The input 3D model is classified by using several features

of the symmetry axis of the input curve skeleton and all symmetry

junctions on that axis. Classification***les are developed from the

general characteristics of each character type. The method also extracts

the anatomical meaning of each curve-skeleton segment. The

template skeleton is then located by using the classification result

without user intervention. Moreover, our classification***les are not

restricted to a specific pose or orientation for the given 3D model,

because all of features used in the***les are local geometry features

which are invariant against pose and orientation.

2 The Proposed Method

The first step involves extracting a curve skeleton from the input

3D model. In this paper, we use our previously developed skeletongrowing

algorithm[Pantuwong and Sugimoto 2010], because this

skeleton-growing process can deliver a clean curve skeleton that

does not require any postprocessing such as a ***ning algorithm.

The proposed method uses the number of symmetry junctions, i.e.,

junctions of legs or arms, as criteria for the classification. However,

the symmetry segments may not always connect to the symmetry