KR101095860B1 - Method for shape deforming feature - Google Patents

Abstract

The shape deformation method of the human body model according to the present invention receives the human body model data composed of the human body scan data or the polygon human body model, generates a joint-skeleton skeleton structure-based key plane, and generates a nurbs surface based on the feature points of the generated key plane. Generating the anatomical model data based on the constructed feature points, generating statistical deformation information on the parameters of the joint points, the key planes and the feature points on the key planes in a plurality of specific motions, and And deforming the human body shape using the information and performing animation.

As described above, the present invention can realistically and naturally transform the human body shape in real time according to the motion capture data capturing the actor's motion using the Nurbs curved human body model based on the feature of the joint-skeleton structure.

3D human body model, Nurbs surface, key plane, feature point, statistical deformation information

Description

Shape deformation method of human body model {METHOD FOR SHAPE DEFORMING FEATURE}

The present invention relates to a deformation of a human body model, and more particularly, a Nurbs surface that approximates key plane curves and key plane curves that approximate a three-dimensional joint-skeleton skeleton structure and key planes between the joints. The human body model that can automatically reproduce various natural motions as shape deformation of the human body model by using statistical deformation information on the key plane constituting the surface from the training data for the specific characteristic motion and the parameters of the feature points on the key plane. It relates to a shape deformation method of.

The present invention is derived from a study conducted as part of the IT original technology development project of the Ministry of Knowledge Economy and the Ministry of Information and Communication Promotion. [Task management number: 2008-F-030-01, Task name: Development of broadcasting convergence type full 3D restoration technology ( Standardization)].

In general, human body models play a very important role in not only computer animation but also 3D games. In order to animate the human body model, the human body model that can be deformed must be created first, and the human body model is an important factor for the quality of the animation.

Conventional human body model generation techniques include a stick model modeling only the initial skeleton, a surface model representing the human body's appearance with curved patches, and a volume composed of a combination of spheres, cylinders, and ellipsoids. Model and so on. However, the models do not express realistic appearances, the shape deformation due to motion is not natural, a large amount of calculation time is required for shape deformation, or manual manipulation by a professional designer or the like. have.

Recently, muscle simulation models reflecting anatomical features and interactive linear blending skinning models based on example data consisting of skeleton and mesh structures have been proposed.

Conventional human body generation model generation method is a model capable of relatively realistic shape deformation, but the difficulty of real-time execution due to the limitation of the calculation speed and the difficulty of production, the accuracy of animation generated according to the accuracy of the pre-made model and the combination of these models There is a problem in which a deformation artefact, such as 'Candy-wrapper', occurs at a major joint.

The present invention performs the animation by deforming the shape of the human model using the human body model data consisting of feature point-based Nurbs surface and statistical deformation information.

The shape deformation method of the human body model according to the present invention receives the human body model data composed of the human body scan data or the polygonal human body model, generates a joint-skeleton skeleton structure-based key plane, and generates the Nurbs surface based on the feature points of the generated key plane. Generating an anatomical model data based on the constructed feature points, an absolute position and orientation with respect to the upper layer joint of the joint, the position of the key plane and at least one specific motion in which the phenomenon of folding, inflation or protrusion occurs at the body connection site; Generating statistical deformation information by using a direction and parameters of the feature points on the key plane, and using the generated feature point-based human body model data and the generated statistical deformation information to deform the human body shape and perform animation. It includes a step.

The present invention uses the Nurbs curved body model based on the feature point of the joint-skeleton structure to automatically deform the human body shape according to the motion capture data that captures the actor's motion, thereby bending, inflating, and projecting the human body during motion. Through the training data on the key joints representing the specific poses and the parameters of the corresponding key planes and feature points, a statistical deformation feature model can be used to provide realistic and natural automatic shape deformation of the feature-based Nurbs surface human model in real time. .

Hereinafter, embodiments of the present invention will be described in detail with reference to the accompanying drawings. In addition, in describing the present invention, when it is determined that the detailed description of the related known configuration or function may obscure the gist of the present invention, the detailed description thereof will be omitted.

1 is a block diagram illustrating an apparatus for explaining a method of deforming a shape of a feature-based human model according to an exemplary embodiment of the present invention.

Referring to FIG. 1, an apparatus for performing shape deformation of a human body model based on a feature point includes an input unit 100 for inputting human body model data consisting of three-dimensional human body scan data and a three-dimensional polygon mesh, and input human body model data. The human body model generator 110 generates the feature-based human body model data consisting of the key plane generated based on the joint-skeleton skeleton structure and the feature-based nurbs surface of the key plane, and the joint and the key plane in various specific motions. Statistical deformation information generation unit 120 for generating statistical deformation information about the parameters of the feature points, and a deformation and animation unit 130 for performing the shape-deformation and animation of the human body model based on the feature point-based human body model and statistical deformation information .

FIG. 2 is a flowchart illustrating a process of generating a feature-based Nurbs curved human body model in FIG. 1.

Referring to FIG. 2, when the mesh human body model, that is, the mesh-based human body model data is input (S200), the human body model generator 110 digitizes the human body model data into motion information of the human body, and thus the joint-skeleton skeleton structure for each part. In other words, as shown in Table 1 below, the user sets a total of m landmarks for each part of the human body (e.g., 71-left-right symmetrical settings) to generate a skeleton structure (S201). . For example, FIG. 5 illustrates a landmark and a skeleton set by a user to generate an initial skeleton structure with a joint-skeleton human model.

division On the body land mark  Location and count arm Shoulder, elbow, cuff 1 each, 20 fingertips (4 parts) 20, 1 fingertip: 24 total Bridge Hips, knees, 1 ankle, 5 toe joints and 1 foot: 9 total body Clavicle Center, Spine, and Pelvis 1 each: 3 in total head Neck, Head: 2 in total

Subsequently, the human body model generating unit 110 uses the spine of the trunk as a root in the generated skeleton structure, and sub-roots the main connection parts (shoulder, wrist, pelvis, ankle, and the like) for each part. As a sub-root, a hierarchical joint structure having a total of n (for example, 68) joints is generated (S202).

Subsequently, the human body model generating unit 110 sets a key plane at a position that can express the appearance well in consideration of the muscles of the body parts between the joints or the joints and joints generated for each body part, and then exists on the plane. The center position is calculated from the set of vertices of the input three-dimensional polygon human body model, the l vertices existing at predetermined intervals, for example, 10 ° intervals, are found at the center point and set as the feature points of the key plane. Interpolation generates a key plane curve (S203).

Then, the human body model generation unit 110 generates a key plane curved surface by Nurbs interpolating the key plane curve for each body part (S204), and generates an appearance of the Nurbs surface-based human body model.

Finally, the human body model generation unit 110 sets a skin vertex connection process for setting a dependency relationship between displacements of the vertices of each key plane curve and the input mesh human body model based on the generated Nurbs surface (S205). A feature point-based Nurbs curved human body model that can deform the shape by controlling the key plane and the feature point through (S206).

3 is a flowchart illustrating a process of generating statistical deformation information in FIG. 1.

Referring to FIG. 3, the statistical deformation information generating unit 120 generates various human-specific motion information in which folds, swells, and protrusions are well represented in the main body connection parts such as shoulders, elbows, hands, necks, knees, and ankles. In operation S301, the 3D human body scan data performing a specific motion or the feature-based Nurbs curved human body model data generated through the process of FIG. 2 in which the professional designer naturally expresses the specific motion is received (S302).

Thereafter, the statistical deformation information generator 120 acquires t pieces of various human body learning data reflecting unique characteristics of the human body such as gender, muscle characteristics, and shape of the human body model used in the specific motion (S303).

Then, the statistical deformation information generation unit 120 hierarchical absolute position and direction parameter data of the upper layer joint of the main joint joint in each specific motion and the parameters of the position, direction and feature points on the key plane of each key plane. In order to acquire the data (S304), that is, repeatedly perform the training data acquired in S303, the parameter data is obtained.

Then, the statistical deformation information generation unit 120 calculates statistical feature information using the obtained parameter data.

In other words, in the t training data obtained in S303, the absolute position, the direction parameter, the position of each key plane, the direction and the feature point positions on the key plane are used for the upper layer joint of the main joint joint. By calculating the statistical feature information (S305).

Finally, the statistical deformation information generation unit 120 generates a mean value of the characteristic information and a variance model of each characteristic information, that is, statistical deformation information generation through Principal Component Analysis (PCA) from the statistical characteristic information calculated in S305 (S306). The statistical deformation information thus generated is used to automatically and naturally transform the feature-based Nurbs surface human body model to be transformed.

FIG. 4 is a flowchart illustrating a shape deformation and animation process of the feature-based Nurbs curved human body model of FIG. 1.

Referring to FIG. 4, the deformation and animation unit 130 receives a feature-based Nurbs curved human body model generated by the human body model generator 110 and a statistical deformation characteristic information model generated by the statistical deformation information generator 120. In addition, motion capture data for a specific motion to be animated is received (S401).

와, 주요연결부위 조인트의 상위 계층 조인트에 대한 절대적 위치, 회전 파라미터를 가중치로 반영한 조인트 파라미터

, 키 평면

의 위치

, 방향

, 크기(키 평면 위의 특징점들

를 반영하는 입력된 통계적 변형 특징정보 모델의 아래의 수학식1로부터 키 평면 및 특징점을 재계산(S404)한다. 예를 들어, 도 6에 도시된 바와 같이, Nurbs 곡면 모델의 팔 부위는 키 평면 및 특징점을 보간하여 생성하는데, 즉 특징점을 보간하여 생성한 키 평면 곡선, 키 평면 곡선을 보간하여 생성한다.Then, the deformation and animation unit 130 performs deformation on the position and rotation of the joint through the motion capture data received from each joint of the input feature-based Nurbs curved human body model (S402), Accordingly, the key plane curve is regenerated by modifying each key plane and the feature point (S403). At this time, the characteristics such as the folding of the skin, the expansion of the muscles, and the protrusion of the human body with respect to a particular pose are weight parameters reflecting the gender, muscle characteristics, shape, etc. of the Nurbs curved body model based on the feature points to be modified.

Joint parameter reflecting the absolute position and rotation parameter of the upper layer joint of the main joint joint as weight

, Key plane

Location of

, direction

, Size (feature points on the key plane

The key plane and the feature point are recalculated from Equation 1 below of the input statistical modified feature information model reflecting S404. For example, as shown in FIG. 6, the arm region of the Nurbs surface model is generated by interpolating key planes and feature points, that is, by interpolating key plane curves and key plane curves generated by interpolating feature points.

는 각각 학습 데이터로부터 얻은 각 키 평면 들의 평균 위치 벡터, 평균 방향 벡터, 평균 크기 벡터들을 나타내고, P는 각각 학습 데이터로부터 얻은 특징정보들의 공분산 행렬의 고유벡터 행렬로 모양 변이의 편차 모드를 규정하고, b는 모델 계수의 벡터로 평균 모양으로부터 변이 정도를 조정한다.here,

Represents an average position vector, an average direction vector, and an average magnitude vector of each of the key planes obtained from the training data, respectively, P is an eigenvector matrix of the covariance matrix of feature information obtained from the training data, respectively, and defines the variation mode of the shape variation. b is a vector of model coefficients that adjust the degree of variation from the mean shape.

The deformation and animation unit 130 interpolates the Nurbs surface based on the recalculated key plane and the feature point to deform the human model shape to fit a specific pose, and repeats this process for all frames to perform the human model animation. (S405). Accordingly, as shown in Figure 7, when a particular pose is taken in the main joint according to the present invention shows a smoothly deformed effect.

The present invention has been limited to the embodiments of the present invention, but it is obvious that the technology of the present invention can be easily modified by those skilled in the art. Such modified embodiments should be included in the technical spirit described in the claims of the present invention.

1 is a block diagram illustrating an apparatus for explaining a method for deforming a shape of a feature-based human model according to an embodiment of the present invention.

FIG. 2 is a flowchart illustrating a process of generating a feature-based Nurbs curved human body model in FIG. 1.

3 is a flowchart illustrating a process of generating statistical deformation information in FIG. 1;

FIG. 4 is a flowchart illustrating a shape deformation and animation process of the feature-based Nurbs curved human body model of FIG. 1.

FIG. 5 is an explanatory diagram illustrating a joint-skeleton human body model according to the present invention, illustrating landmarks and skeletons set by a user to generate an initial skeleton structure;

FIG. 6 is a diagram showing an arm plane of a key plane, a key plane curve generated by interpolating a feature point, and a feature point, and a Nurbs surface model generated by interpolating a key plane curve according to the present invention.

7 is a view showing a smoothly deformed effect when a particular pose is taken in the main joint according to the present invention.

Claims (8)

Generating a joint-skeleton skeleton structure-based key plane by receiving the human body scan data or the human body model data composed of the polygonal human body model, and generating the feature-point-based human body model data composed of the feature-based Nurbs surface of the generated key plane; , By using the parameters of the absolute position and orientation of the joint's upper layer joint, the position and orientation of the key plane and the feature points on the key plane in at least one specific motion in which the phenomenon of folding, inflation or protrusion at the body connection occurs. Generating statistical transformation information, Deforming the human body model shape and performing animation by using the generated feature point-based human body model data and the generated statistical deformation information. Shape deformation method of the human body model comprising a. The method of claim 1, Generating the human body model data based on the feature points, Generating the joint-skeleton skeleton structure for each part by digitizing the human body scan data or the polygon human body model with motion information of the human body; Generating the human body model data based on the feature points based on the generated joint-skeleton skeleton structure Shape deformation method of the human body model comprising a. The method of claim 2, Generating the joint-skeleton skeleton structure, Receiving m human body model data and generating m landmarks and a skeleton structure for each part of the human body; In the generated skeleton structure, the spine of the trunk is the root, and the connection part of each part is the sub-root.

A hierarchical joint structure having two joints Shape deformation method of the human body model comprising a. The method of claim 2, Generating the human body model data based on the feature points, Setting a key plane at the selected position after selecting a position in consideration of the muscles of the joints or the joints and the joints generated for each region; Calculating a center position in the set of vertices of the polygonal human body model existing on the set key plane; Present at predetermined intervals at the calculated center position

Finding and setting the number of vertices as feature points of the key plane, and generating key plane curves by B-spline interpolation of the feature points; Nurbs interpolating the generated key plane curve for each part to generate a key plane curved surface; Setting a dependency relationship between displacements of each of the key plane curves and the vertices of the input mesh human body model based on the generated key plane curves; Generating the human body model data based on the feature points that can deform the shape by controlling the key plane and the feature points based on the set dependency; Shape deformation method of the human body model comprising a. The method of claim 1, Generating the statistical deformation information, Setting a specific motion in which the fold, inflation, or protrusion occurs in the body connection part; Acquiring the hierarchical absolute position and direction of the joints at the main connection part of the human body model during the set specific motion, the position and direction of each of the key planes, and the parameter data of the feature points on the key plane; Generating the statistical deformation information using the obtained parameter data Shape deformation method of the human body model comprising a. The method of claim 5, Acquiring the parameter data, Acquiring a plurality of human body learning data reflecting unique characteristics of the human body in the specific motion; Acquiring the parameter data of the hierarchical absolute position and direction of the main connection joints of the acquired learning data, the position and direction of each of the key planes, and the feature points on the key planes; Shape deformation method of the human body model comprising a. The method of claim 6, Generating the statistical deformation information, Calculating characteristic information using a change of parameter data obtained from each learning data; Generating the statistical deformation information which is a variance model of the average value of the feature information and the feature information through principal component analysis of the calculated feature information Shape deformation method of the human body model comprising a. The method of claim 1, Performing the animation, Receiving the generated human body model data based on the feature points, the generated statistical deformation information, and motion capture data for the specific motion; Performing deformation on the position and rotation of the joint through the motion capture data received from each joint of the generated feature point-based human body model data; Regenerating a key plane curve by modifying each key plane and a feature point according to each modified joint; Performing a human body model animation by repeating the human body model data by interpolating the human body model data based on the modified key plane and the characteristic point and repeating the human body model shape deformation for every frame. Shape deformation method of the human body model comprising a.

Images